A Look at the Legal Environment for Driverless Vehicles (2016)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

17 Whether similar echoes follow from the policies adopted toward more recent technologies, such as computers, remains to be seen. What do these principles mean for the legal envi- ronment for driverless vehicles? Above all else, they point toward an evolving policy response to these devices. Policymaking probably will begin with rudi- mentary measures and become more complex and far-reaching over time. At first, the aspects of driver- less vehicles regarded as most suitable for regula- tion will be defined largely, if not entirely, by refer- ence to the law that surrounds conventional vehicles. But as driverless vehicles grow increasingly sophis- ticated and common, more and more novel issues will arise that will require innovative and thought- ful responses from policymakers. Some of these poli- cies eventually may produce far-reaching changes in the ambient law. At the same time, past experience also suggests that the policymaking path for new technologies can be unpredictable. As earlier episodes establish, a sin- gle incident may cause lawmakers to step in to address the perceived risks associated with a novel technology. Therefore, forecasts regarding the “likely” or “optimal” policy responses to driverless vehicles should be ventured with caution, and with an appre- ciation that alternative approaches may prevail. With that, this report will conclude its discussion of other, past technologies and proceed to the charac- teristics and potential uses of driverless vehicles. III. Characteristics and Technologies of Driverless Vehicles Driverless vehicles are motor vehicles in which internal vehicle systems, instead of a human driver, operate all functions as the vehicle moves on public roadways. They can take the form of passenger cars, large or small trucks, buses, or other modes of motor- ized ground transportation. They may transport either cargo or human passengers, or both, or neither. Varied terminology regarding motor vehicles without human drivers complicates policymaking about driverless vehicles. These vehicles are some- times referred to as “entirely self-driving,” as “fully autonomous,” or even as “completely automated,” in addition to being more aptly called “driverless vehi- cles.”178 The negative attribute of being driverless echoes the “horseless carriages” name for human- driven automobiles over a hundred years ago. Both phrases suggest a comfortable transformation from something familiar into something new that is, in fact, transformative. Some of the same technologies used in driverless vehicles provide automated features in conventional vehicles. But automated vehicles are not necessarily driverless. Already available automated, semi- autonomous, or self-driving technologies assist human drivers who control all or some of the vehi- cles’ operations. Familiar automated technologies currently assist drivers with specific vehicle func- tions, such as braking or parking, but continue to need a human driver to control general vehicle oper- ations. Moreover, international conventions179 and state laws expressly require a driver to be in control of a motor vehicle operating on public roads.180 Currently available varieties of automated, self- driving, semi-autonomous, or connected features on conventional vehicles will provide experience with the application of some of the types of technologies that are also applied in driverless vehicles. On-road Colorado held that “driverless,” as applied to cars, was even then generic and therefore did not infringe any trademark. Id. at 270. Many of the earliest driverless car cases come from Texas, where in 1918 the City of San Antonio adopted “An ordinance for the licensing and regulation of driverless automobiles, hired or leased to the public for use on or over the streets or thoroughfares of the city of San Antonio.” City of San Antonio v. Besteiro, 209 S.W. 472, 472 (Tex. Ct. Civ. App. 1919), upheld this ordinance, which required registra- tion and financial responsibility. Id. at 473–74. Other state courts, as well as the United States Supreme Court, simi- larly upheld driverless-car regulations. E.g., Hodge Drive-It- Yourself Co v. City of Cincinnati, 284 U.S. 335, 52 S. Ct. 144, 76 L. Ed. 323 (1932). 179 Convention on Road Traffic, Geneva, Sept. 19, 1949, 3 U.S.T. 3008, 125 U.N.T.S. 3 and Vienna Convention on the Law of Treaties, art. 31, May 23, 1969, 1155 U.N.T.S. 331, 8 I.L.M. 679. Article 8, Sections 1 and 5 of the Vienna Convention require that “[e]very moving vehicle or combi- nation of vehicles shall have a [person] driver,” and “[e]very driver shall at all times be able to control his vehicle.” Section 1 of Article 13 further requires, “Every driver of a vehicle shall in all circumstances have his vehicle under control so as to be able to exercise due and proper care and to be at all times in a position to perform all manoeuvres required of him.” 180 Laws in some states that govern driverless or run- away vehicles impose liability on the owner of a driverless vehicle that causes damage or injury. E.g., CAl veh. Code § 16001 (2015). Most states have laws that require a licensed driver be in control of motor vehicles. Self- Driving Cars and Insurance, ins. info. insT. (Feb. 2015), http://www.iii.org/issue-update/self-driving-cars-and- insurance. Still, vehicle automation that assists drivers is generally legal in the United States. Bryant Walker Smith, Automated Vehicles Are Probably Legal in the United States, 1 Tex. A&m l. rev. 411 (2014). 178 Early in the 20th century, “driverless automobiles” referred to a regulated business now known as the rental car business. Instead of hiring a motor vehicle with a driver, a person could rent a “Driverless Car” from a “business of renting and hiring automobiles and motor vehicles without a driver.” Driverless Car Co. v. Glessner-Thornberry Driv- erless Car Co., 83 Colo. 262, 264, 264 P. 653, 654 (1928). In this trademark infringement case, the Supreme Court of

18 performance of these limited self-driving features, or autonomous operational modes, or automated opera- tions will contribute data to inform legal policy deci- sions about vehicles that operate entirely without human drivers. Still, legal and policy issues posed by driverless vehicles operating without human drivers are very different from issues posed by vehicles that have not entirely obviated a human operator. A. Distinctive Characteristics of Driverless Vehicles Driverless vehicles offer a means of roadway transportation for both goods and people that oper- ates and controls its own operations and movements. Driverless vehicles are expected to be safer, more efficient, and more environmentally benign than conventional driver-operated vehicles.181 They are expected to save thousands of lives and millions of dollars in avoided damage and waste. Existing motor vehicle regulatory requirements (such as extensive passive safety equipment, pollution control devices, and the like) may cause the earliest driverless vehi- cles outwardly to appear similar to conventional vehicles. Ultimately, driverless vehicles will probably look quite different from human-driven vehicles. Driverless vehicles’ crash-avoidance capabilities should obviate the need for robust passive safety fea- tures, such as heavy materials, bumpers, air bags, or even full-visibility windshields. However, before such physical changes can occur, a great deal of legal and regulatory change will need to take place. 1. Autonomous, Automated, and Self-Driving Features Various applications of automated vehicle sys- tems—from electronic stability control to automatic lane keeping, parking, and braking systems—enable vehicles to perform specific tasks without human intervention. Recently introduced automated sys- tems that control some or all vehicle operations for part of a journey or in a specific roadway environ- ment are also available. However, for now, human drivers remain in overall control, particularly in emergencies. General Motors’ “Super Cruise”182 and Tesla’s promised “Autopilot”183 are brand-associated features advertised as making vehicles autonomous or self-driving. However, even such highly automated vehicle functions and self-driving modes do not enable cars having such automated features to be driverless, in the sense of dispensing entirely with a human driver. A human driver remains both legally and practically required to be present and capable of positive operational control of these vehicles. In con- trast, driverless vehicles operate without any human control and therefore pose legal and policy issues dif- ferent from those posed by vehicles that continue to rely on the presence of a human driver. 2. Levels of Automation For many advanced automotive system develop- ers, “autonomous” became sufficiently ambiguous that standard-setting and regulatory bodies avoid using it in favor of “a range of vehicle automation.” Increasing degrees of vehicle automation, in an inverse relationship with human control, seems helpful in describing the increasingly sophisticated stages of vehicle automation technologies. In fact, there are two separate versions of vehicle automation levels. For both of them, driverless vehi- cles are at the highest level of full automation,― meaning that the vehicle is in complete control of all driving functions at all times. In 2013, the National Highway Traffic Safety Administration (NHTSA) suggested vehicle automation levels in the agency’s “Preliminary Statement of Policy Concerning Auto- mated Vehicles.”184 In January 2014, Society of Automotive Engineers (SAE) International sug- gested a slightly different “Taxonomy and Definitions for Terms Related to On-Road Motor Vehicle Auto- mated Driving Systems” as SAE Standard J3016.185 The two competing sets of categories, or levels, are: SAE Automation Levels 0—No Automation 1—Driver Assistance 2—Partial Automation 3—Conditional Automation 4—High Automation 5—Full Automation [Driverless] NHTSA Automation Levels Level 0—No Automation Level 1—Function-specific Automation 181 See JAmes m. Anderson eT Al., AUTonomoUs vehiCle TeChnology: A gUide for PoliCymAkers 2 (2014), http:// www.rand.org/content/dam/rand/pubs/research_reports/ RR400/RR443-1/RAND_RR443-1.pdf (hereinafter “RAND rePorT”), at 9–40. 182 Jerry Hirsch, GM will introduce hands-free, foot-free driving in 2017 Cadillac, L.A. Times (Sept. 7, 2014), http:// www.latimes.com/business/autos/la-fi-hy-gm-cadillac- super-cruise-20140907-story.html. 183 Christopher DeMorro, Elon Musk: Tesla Capable of “90% Autopilot” By Next Year, CleAnTeChniCA (Oct. 7, 2014), http://cleantechnica.com/2014/10/07/elon-musk-tesla- capable-90-autopilot-next-year/. 184 nAT’l highWAy TrAffiC sAfeTy Admin., U.s. deP’T of TrAnsP., PreliminAry sTATemenT of PoliCy ConCerning AU- TomATed vehiCles (2013) (hereinafter “NHTSA PreliminAry sTATemenT”), http://www.nhtsa.gov/staticfiles/rulemaking/ pdf/Automated_Vehicles_Policy.pdf. 185 SAE International, Taxonomy and Definitions for Terms Related to On-Road Motor Vehicle Automated Driv- ing Systems, SAE International Standard J3016, issued 2014-01-16. Summary table available at http://www.sae. org/misc/pdfs/automated_driving.pdf.

19 Level 2—Combined Function Automation Level 3—Limited Self-Driving Automation Level 4—Full Self-Driving Automation [Driverless] It is noteworthy that driverless vehicles occupy the highest level of automation in both systems.186 Under NHTSA categories, currently available vehicle auto- mation technologies are at level 2, and are rapidly moving into level 3, but are not yet close to the driver- less top NHTSA automation level 4—completely driv- erless operation. Similarly, existing vehicle automa- tion is currently between SAE levels 2 and 3. Federal and state regulators in the United States generally refer to NHTSA vehicle automation levels. Vehicle manufacturers often use the SAE categories, which are similar to vehicle automation categories used in Europe. This report uses the NHTSA levels of vehicle automation as reference points. 3. Consumer Acceptance Highly automated, but not completely driverless, vehicles that retain a human driver in the control/ responsibility loop may be sufficient for many vehi- cle purchasers for many years. Particularly with regard to passenger vehicles, there may turn out to be a consumer-market “stickiness” at levels of vehi- cle automation that are less than fully driverless. Consumers may be satisfied with highly automated driver assistance, and on-demand self-driving modes in certain circumstances, but still prefer to retain control over their own vehicles. Some generations of car buyers, for whom a driver’s license typically crowned their adolescence, may not be as eager to leave car operation to the car. Other poten- tial driverless vehicle buyers express feelings of inse- curity if vehicle control is not in the hands of a human driver. In addition to anxieties about an “uncontrolled” driverless vehicle, some consumer skepticism about driverless passenger cars undoubtedly reflects legal uncertainties, as well as concerns about safety and financial risks. A generational change in car purchas- ers to those whose personal attachment to vehicle driving is more attenuated, as well as changes in legal regimes, are likely to be required before driverless vehicles become widespread consumer choices over highly automated driver-assisting vehicles. In addition to psychological reluctance to relinquish control over personal mobility, legal consequences of having no human driver in control, or potential con- trol, of a passenger car are pervasive. In some areas of law, such as vehicle regulation and insurance, driver- less vehicles may require entirely new legal rules. Gradual legal-system adaptation to vehicle automa- tion—such as acceptance of limited self-driving modes or part-time driver passivity—would provide valuable experience in guiding legal changes that will be needed before large numbers of completely driverless vehicles operate on United States roadways. Although driverless vehicle market penetration estimates187 vary a great deal, some type of fully driverless vehicles (passenger cars or trucks) are expected to be commercially available by around 2025.188 Some estimates indicate that driverless vehicles will not become standard until the 2050s, because of the rather slow vehicle replacement rate. By 2050, driverless vehicles are expected to account for over half of vehicle travel.189 B. Driverless Vehicle Technologies Interaction among many types of technologies will enable driverless vehicles to operate on public roads without being operated by human drivers. The complexities of these technical systems will present unusual challenges to courts and legislatures tasked with creating and applying legal rules regarding driverless vehicles. The following description of driverless vehicle technologies is homocentric. It starts with the inter- face between a human user and a driverless vehicle, then considers several types of data input technolo- gies, as well as automated controls over vehicle func- tions and the artificial intelligence technologies that integrate data input and determine when and how to activate automated vehicle controls. This discussion will consider five groups of technologies that combine to operate a driverless vehicle: 186 Driverless vehicles are classified as NHTSA Level 4–Full Self-Driving Automation: The vehicle is designed to perform all safety-critical driving functions and monitor roadway conditions for an entire trip. Such a design anticipates that the driver will initially provide destination or navigation input, but is not expected to be available for control at any time during the trip. This includes both occupied and unoccupied vehicles. By design, safe operation rests solely on the automated vehicle system. Similarly, driverless vehicles occupy SAE Level 5: Full Automation, which refers to driverless vehicles as involv- ing “the full-time performance by an automated driving system of all aspects of the dynamic driving task under all roadway and environmental conditions that can be man- aged by a human driver.” 187 See, e.g., ernsT & yoUng, dePloying AUTonomoUs vehi- Cles: CommerCiAl ConsiderATions And UrbAn mobiliTy sCe- nArios (2014), http://www.ey.com/Publication/vwLUAssets/ EY-Deploying-autonomous-vehicles-30May14/$FILE/EY- Deploying-autonomous-vehicles-30May14.pdf. 188 xAvier mosqUeT, ThomAs dAUner, nikolAUs lAng, miChAel rübmAn, mei PAChTler, rAkishiTA AgrAWAl & floriAn sChmeig, revolUTion in The driver’s seAT: The roAd To AUTonomoUs vehiCles 7 (2015) (hereinafter “bosTon ConsUlTing groUP rePorT”). 189 Todd liTmAn, AUTonomoUs vehiCle imPlemenTA- Tion PrediCTions imPliCATions for TrAnsPorT PlAnning 12 (2015), http://www.vtpi.org/avip.pdf.

20 1. Human-vehicle interface. 2. Sensors that provide data about internal operation of the vehicle and its parts. 3. Sensors that provide location and real-time external roadway environment data. 4. Automated controls over vehicle functions and operation. 5. Artificial intelligence that integrates in- vehicle operational data with external roadway data and activates automated vehicle controls. Each of these groups of vehicle technologies pres- ents multiple challenges to the legal system. Integrat- ing these technologies to operate a driverless vehicle poses additional technical and legal challenges. 1. Human-Vehicle Interface The points at which a human user interacts with a driverless vehicle will be crucial in determining legal responsibility. These interaction points, also called HMI (human-machine-interface), will be the loci of human choices regarding driverless vehicles. It is likely that, at least at first, driverless vehicles will involve a simple HMI that provides no choices other than to use the driverless vehicle or not to use it and to select the vehicle’s destination. That interface may be biometric (a fingerprint reader or speaker recogni- tion) or may take the form of a fob, a push button, password entry, or other on-off control. For security purposes, at least two means of verification (access factors) are likely to be required. In some jurisdic- tions, testing regulations provide that a human per- son turning on a driverless vehicle makes that human the legal “operator” of the driverless vehicle.190 It seems likely that manufacturers will program early driverless vehicles in standard ways. At some point in their development, driverless vehicles may provide more complex interactions between humans and driverless vehicles than simply activating the vehicle and setting its destination. Experimental driverless vehicles are generally programed to obey all traffic signs and rules, to avoid crashes, to stay in their appropriate lanes, etc. It is likely that driverless car users will want to be able at least to change the vehicle’s destination during a journey, when the purpose for a trip changes. These and other programming choices could cause a driverless vehicle to operate in ways that uniquely respond to individual users’ person- alities and preferences. Unless legal requirements restrict the program- ming of production versions to a standard driving mode, driverless vehicles could offer a menu of alter- native programming that would provide choices among different driving styles. For example, a driv- erless car user could select among such driving styles as “aggressive driving,” “slow driving,” “scenic routes,” or “arrival by <specified time> at all costs.” Although default driverless vehicle programming is expected to direct “safest and most time- and fuel-efficient, while obeying all traffic laws,” pro- gramming alternatives are inevitable. These user alternatives would provide choices among specific ways in which the driverless vehicle functions, based on different coding menus provided by programmers. Absent legal restrictions on permitted and unper- mitted driverless vehicle computer code, experimen- tation by programmers and hackers seems inevitable. Laws, regulations, or purchase contracts forbidding particular types of driverless vehicle programming or reprogramming would be difficult to write and to enforce, but may be on the horizon. Increased human choices regarding how a driver- less vehicle operates will generate human responsi- bility, including legal responsibility, for choosing more or less risky driverless vehicle operating modalities. Future driverless vehicle laws or regula- tions may well require specified programming limi- tations (e.g., always abide by all traffic laws and sig- nals) or provide safe harbors that limit legal liability of driverless vehicle users if specified programming is used. However, at present such laws or regula- tions do not exist. Although ethical obligations of driverless vehicle programmers have been dis- cussed, legal rules governing such matters have not been drafted. 2. Sensors Providing Internal Vehicle Operation Data Sensors that detect and process the operation of various vehicle parts, such as the brakes, transmis- sion, steering, throttle, tires, and the like are already embedded in all modern vehicles.191 Thousands of sensor microprocessors communicate over the CAN bus (under ISO 11898 standards) for vehicle coordi- nation, diagnostic, and other purposes.192 The capacities and configurations of these sensors are typically proprietary information closely held by 190 For example, under Florida law, “a person shall be deemed to be the operator of a driverless vehicle operating in driverless mode when the person causes the vehicle’s driverless technology to engage, regardless of whether the person is physically present in the vehicle while the vehicle is operating in driverless mode.” flA. sTAT. § 316.85 (2014). 191 Robert N. Charette, This Car Runs on Code, IEEE sPeCTrUm (Feb. 1, 2009), http://spectrum.ieee.org/ transportation/systems/this-car-runs-on-code. 192 Steve Corrigan, Introduction to the Controller Area Network (CAN) (July 2008), http://www.ti.com/lit/an/ sloa101a/sloa101a.pdf. ISO 11898-1:2003, “Road vehicles —Controller area network (CAN)—Part 1: Data link layer and physical signaling” is available through http://www. iso.org/iso/catalogue_detail.htm?csnumber=33422.

21 vehicle manufacturers.193 Because these sensors func- tion as evaluators of the internal mechanical opera- tions of a vehicle and its parts, the information they generate can have significant legal consequences in terms of causing, diagnosing, or isolating vehicle mal- functions. These internally facing sensors also provide points of access for intruders to insert malicious code that could misdirect or even take control of a driver- less vehicle. Recent reports about remote carhack- ing194 illustrate this security vulnerability, which also is discussed in Section VII. 3. Sensors Providing Location and External Roadway Environment Data Global positioning systems (GPS) that provide real-time location information are a nearly univer- sal feature of experimental driverless cars. However, because the resolution of ordinary GPS signals is only accurate to a level of 3.5 meters, augmentation (such as through differential GPS) is required to geographically locate a driverless vehicle within a few centimeters.195 In addition to GPS, precise loca- tion information mostly comes from dynamic digital mapping. Experimental driverless vehicles gener- ate, as well as use, digital maps of roadways. Precise, real-time mapping, tracking, and other “environ- mental awareness” technologies used by driverless vehicles are essential to safe vehicle operation.196 As a result, most driverless cars will routinely receive as well as generate mapping updates at frequent intervals. It is possible that this dynamic mapping data could be provided as cloud-sourced driverless vehicle roadway data. Experimental driverless vehicles depend on out- ward-facing sensors that collect real-time data about what is happening in the immediate and lon- ger-range roadway environment through which a driverless vehicle is moving. For example, Google has a specific patented sensor process for noting and reacting to such unexpected events as cattle cross- ing the road.197 Indeed, driverless vehicle developers have invented a variety of different types of sen- sors.198 Multiple forms of radar, LIDAR,199 infrared, sonar, and optics (digital cameras) combine to pro- vide a detailed and robust “picture” of the immedi- ate and farther away roadway environment. These multiple sensors operate as redundant sources that provide both static (a curb or pothole) and dynamic (bicyclist alongside) roadway data. Poor weather conditions interfere with many sensors that require line-of-sight. As a result, in some climates and cir- cumstances, even redundant arrays of multiple sen- sors may fail to provide adequate roadway data for driverless cars.200 To supplement these sensors, wireless communi- cations are expected also to supply roadway situa- tional information (e.g., movement of nearby vehi- cles) to driverless vehicles, particularly in circumstances where weather compromises visibil- ity. These wireless communications technologies are not sensors, but they can provide vital data about a driverless vehicle’s dynamic roadway environment. 193 Much of this proprietary data is protected as trade secrets. The software that operates vehicle systems is copy- righted. As discussed in Section VII.C, infra, at note 546, vehicle manufacturers are objecting to a U.S. Copyright Office proposal to exempt decompiling and modifying this software from being considered illegal tampering with anti- circumvention measures that protect digital barriers against copyright infringement. 194 Andy Greenberg, Hackers Remotely Kill a Jeep on the Highway—With Me in It, Wired (July 21, 2015), http:// www.wired.com/2015/07/hackers-remotely-kill-jeep- highway/. As a consequence, over a million vehicles were subject to a product recall in which software updates were sent to their owners. Andy Greenberg, After Jeep Hack, Chrysler Recalls 1.4M Vehicles for Bug Fix, Wired (July 24, 2015), http://www.wired.com/2015/07/jeep-hack-chrysler- recalls-1-4m-vehicles-bug-fix/. 195 William Messner, sAe inTernATionAl & AUvsi, AUTon- omoUs TeChnologies: APPliCATions ThAT mATTer 11, 108–89 (2014). 196 Greg Miller, Driverless Cars Will Require a Totally New Kind of Map, Wired (Dec. 15, 2014), http://www. wired.com/2014/12/nokia-here-autonomous-car-maps/; Vince Bond Jr., Up-to-the-minute maps will be critical for driverless cars, AUTomoTive neWs (Sept. 13, 2014), http:// www.autonews.com/article/20140913/OEM06/309159962/ up-to-the-minute-maps-will-be-critical-for-driverless- cars. An interesting technical description of how driverless vehicles create maps is Pierre Lamon, Cyrill Stachniss, Rudolph Triebel, Patrick Pfaff, Christian Plagemann, Giorgio Grisetti, Sascha Kolski, Wolfram Burgard & Roland Siegwart, Mapping with an Autonomous Car (2006), http://www2.informatik.uni-freiburg.de/~grisetti/pdf/ lamon06iros.pdf. 197 U.S. Patent 8,996,224, issued Mar. 31, 2015. See also Mary Beth Griggs, Google’s Driverless Cars Are Learning How To Avoid Cows, PoPUlAr sCienCe (Apr. 7, 2015), http:// www.popsci.com/google-making-sure-its-driverless-cars- wont-crash-cows-0. 198 Greg Kogut, Sensors, in driverless TeChnologies, supra note 195, at 1–13; Ryan Whitwam, How Google’s self-driving cars detect and avoid obstacles, exTremeTeCh (Sept. 8, 2014), http://www.extremetech.com/extreme/ 189486-how-googles-self-driving-cars-detect-and-avoid- obstacles. 199 “LIDAR—Light Detection and Ranging—is a remote sensing method used to examine the surface of the Earth.” What Is LIDAR?, nAT’l oCeAniC And ATmosPheriC Admin., http://oceanservice.noaa.gov/facts/lidar.html (last visited Sept. 20, 2015). 200 Doron Levin, The cold, hard truth about driver- less vehicles and weather, forTUne (Feb. 20, 2015), http://fortune.com/2015/02/02/autonomous-driving- bad-weather/.

22 Section III.C.2.a discusses the potential application of connected vehicle technologies in driverless vehi- cles. In addition, beacon technologies, or pavement- embedded signals, may be developed to transmit to driverless vehicles information from signs and warnings about potential roadway hazards such as low bridges, tight curves, or lane closures. 4. Automated Controls over Vehicle Functions and Operation In a driverless vehicle, control over vehicle opera- tion is automated through networks of actuator microprocessors (sometimes called ECUs, for “elec- tronic control units”) triggered by the vehicle’s arti- ficial intelligence. So far, automated controls in con- ventional vehicles appear to have been remarkably reliable in accomplishing specific vehicle operations from anti-lock brakes to electronic stability control. However, some automated vehicle controls appear to have proved more reliable than others. For example, automatic lane-keeping controls201 seem to be less reliable than electronic stability control.202 Media reports about technical experiments enabling remote access to automated vehicle con- trols have eroded public confidence in automated vehicle controls. Such vulnerabilities, associated with car-hacking, present legal as well as technical challenges for driverless vehicles. Indeed, automated controls have proved to be the most vulnerable aspect of vehicle automation to car hacking.203 The security aspects of automated controls in driverless vehicles will be discussed further in Section VII.C. 5. Artificial Intelligence Driverless vehicles rely on highly sophisticated artificial intelligence to integrate and analyze inter- nal vehicle operational data and roadway sensor data and then to determine which automated con- trols to activate. This machine ability to control all vehicle operations distinguishes driverless vehicles from other automated technologies that either assist or warn human drivers. Driverless vehicle artificial intelligence inte- grates internal vehicle operational and external roadway environment inputs as described above. It is likely that driverless vehicle artificial intelligence will be functionally distributed across multiple parts of a vehicle’s decision and control systems, rather than being located in a single central processing unit. It also will be self-learning in the sense that the algorithms used in operating a vehicle modify themselves over time in response to previous opera- tions, new information, and feedback. Self-learning algorithms are characterized by their dynamic adaptability. Rather than robotically carrying out static programming directions, driverless vehicles analyze data, model it, and make data-driven pre- dictions and decisions, such as actuating vehicle controls.204 Actuated controls simultaneously pro- vide feedback data to various parts of the system. So far, sufficient computational power to manage driverless vehicle data integration, analysis, and activation appears to be available and at necessary analytic speed.205 However, capacities for rapid data fusion and control architecture are not unlimited. In particular, the computational demands of advanced security systems needed to protect driverless vehi- cles from external threats may drain resources and slow analytic functioning in driverless vehicles. A driverless vehicle’s artificial intelligence is tasked with performing vehicle management and guidance functions otherwise performed by a human driver. That artificial intelligence has to be at least as accu- rate and reliable as human intelligence engaged in the same types of operations.206 At present, the legal system does not specifically regulate any of the parameters in which driverless vehicle artificial intelligence will be permitted to operate. Because artificial intelligence decisions have consequences in terms of safety, economic, and environmental impacts, this aspect of driverless cars is likely to be subject to extensive legal regula- tion that is not yet in existence. C. Connected Vehicle Technologies Various types of connected vehicle technologies (wireless communications) may provide inputs for driverless vehicle operation. As of mid-2015, it 204 Self-learning artificial intelligence is a type of machine learning developed in computer science. See gen- erally Stuart Russell and Peter Norvig, ArTifiCiAl inTelli- genCe: A modern APProACh (2013). 205 Aaron Dubrow, Researchers improve artificial intel- ligence algorithms for semi-driverless vehicles, PhysiCs.org (Feb. 3, 2015), http://phys.org/print342167574.html. 206 Some of the most sophisticated applications of self- learning artificial intelligence come from DeepMind, now owned by Google, Inc. See Hannah Devlin, Google develops computer program capable of learning tasks independently, The gUArdiAn (Feb. 25, 2015), http://www.theguardian. com/technology/2015/feb/25/google-develops-computer- program-capable-of-learning-tasks-independently. 201 Chad Kirchner, Lane Keeping Assist Explained, mo- Tor revieW (Feb. 17, 2014), http://motorreview.com/lane- keeping-assist-explained/. It may be that lane markings are insufficiently standardized and maintained for the technology to operate properly. 202 Electronic Stability Control, NHTSA, http://www. safercar.gov/Vehicle+Shoppers/Rollover/Electronic+ Stability+Control (last visited Sept. 20, 2015). 203 ChArlie miller & Chris vAlAsek, remoTe exPloiTA- Tion of An UnAlTered PAssenger vehiCle (2015), http:// illmatics.com/Remote%20Car%20Hacking.pdf.

23 remains undetermined which types of vehicle com- munications will be integrated into driverless vehi- cles. Driverless vehicles are not technically required to be connected vehicles. But they probably will be. 1. Vehicle Connectivity Wireless communications systems connect vehi- cles with other vehicles or with other receivers located in or near the roadway around them. Various technologies provide this vehicle connectivity. Wire- less technologies are already embedded in most late- model conventional cars. Cellular wireless connec- tions offer phones, Internet access, information, and entertainment to moving vehicles. A NHTSA requirement of dedicated short range communica- tions (DSRC) transceivers as safety equipment on all new passenger vehicles and light trucks is expected to be proposed by the end of 2015.207 All of these systems could provide data useful in the oper- ation of driverless vehicles. Ambiguous terminology frequently confuses pol- icy decisions regarding vehicle connectivity. In the United States, “telematics” (also “automotive telematics” or “mobile telematics”) can refer generi- cally to any form of wireless communication to or from vehicles, i.e., vehicles connected to the outside world over various types of wireless connections. For some automotive analysts, such as the consulting firm Gartner, only communications from vehicles are telematics: “Telematics refers to the use of wire- less devices and ‘black box’ technologies to transmit data in real time back to an organization. Typically, it’s used in the context of automobiles, whereby installed or after-factory boxes collect and transmit data on vehicle use, maintenance requirements or automotive servicing.”208 “Mobile telematics” can refer, even more narrowly, to connections between an automobile’s computer systems and embedded wireless communications systems that transmit vehicle operation data to the vehicle’s manufacturer or insurer. Under whatever name, telematics or var- ious other vehicle communications technologies may be used in driverless vehicles. On the other hand, because driverless vehicles can operate without wireless communications functions, driverless vehi- cles need not have any such connectivity. Indeed, some experimental driverless vehicles have been deliberately designed not to connect with external sources of information for their operation. For example, the vehicles involved in the DARPA 2004, 2005, and 2007 Grand and Urban Challenges were not permitted to use externally communicated information for vehicle operation.209 One way to characterize vehicles that rely only on data gener- ated within the vehicle, without wireless connec- tions, is to describe them as “self-contained,” in con- trast with wirelessly connected vehicles, described as “interconnected.”210 What is not known at this time is whether future driverless cars will be interconnected210 through reliance on wirelessly communicated data for vehicle operations. 2. USDOT Connected Vehicle Program The United States Department of Transportation (USDOT) has an elaborate Connected Vehicle Pro- gram.211 This program is divided into two parts: 1) connected vehicle safety systems through which vehicles transmit and receive vehicle operation data over DSRC transceivers, and 2) connected vehicle mobility applications, which provide information, entertainment, and other communications over var- ious commercial wireless networks.212 USDOT Connected Vehicle Program research has also considered “multi-modal” systems combining DSRC safety data and other wireless vehicle com- munications platforms into a “Core System” that would connect vehicles with off-road transportation management systems and other vehicle data users.213 Only a concept of operations has been pre- pared for such a combined Core System. USDOT’s bifurcated Connected Vehicle Program divides DSRC wireless vehicle communications 207 Transportation Sec. Foxx announces steps to ac- celerate road safety innovation DOT 49-15, NHTSA (May 13, 2015), http://www.nhtsa.gov/About+NHTSA/ Press+Releases/2015/nhtsa-will-accelerate-v2v-efforts. 208 Gartner IT Glossary “Telematics,” gArTner (2013), http://www.gartner.com/it-glossary/telematics. 209 See DARPA Urban Challenge, DARPA, http:// archive.darpa.mil/grandchallenge/ (last visited Sept. 20, 2015). 210 Dorothy J. Glancy, Privacy in Autonomous Vehicles, 52 sAnTA ClArA l. rev. 1171, 1176–78 (2012). 211 The Intelligent Transportation Systems Joint Pro- gram Office (ITS JPO) manages research aspects of the connected vehicle program within the Office of the Assis- tant Secretary for Research and Technology. The Office maintains a Web site on this subject at http://www.its.dot. gov/connected_vehicle/connected_vehicle_research.htm. In addition, the Federal Highway Administration Office of Operations also conducts connected vehicle research. Their connected-vehicle Web site appears at http://ops. fhwa.dot.gov/travelinfo/infostructure/aboutinfo.htm. 212 Christopher Hill, modUle 13: ConneCTed vehiCles 1–2 (2013), http://www.pcb.its.dot.gov/eprimer/documents/ module13.pdf (“[I]n summary, safety-related systems in the connected vehicle environment will likely be based on dedicated short range communications (DSRC). Non- safety applications may be based on different types of wireless technology.”). 213 Connected Vehicle Core System Baseline Documen- tation, offiCe of The AssisTAnT seC’y for reseArCh And TeCh., U.s. deP’T of TrAnsP. (Sept. 16, 2015), http://www. its.dot.gov/connected_vehicle/connected_vehicle_ coresystems.htm.

24 technologies that use closed ad hoc networks from commercial mobile wireless applications. Indeed, the wireless technologies involved are both techni- cally and legally distinct. For example, cybersecurity vulnerabilities posed by DSRC connected vehicles are markedly different from those posed by commer- cial wireless mobile applications. Referring to both categories as “connected vehi- cles” obscures important technical differences that affect legal policy determinations. Indeed, the legal ramifications of safety-oriented DSRC vehicle com- munications are unlike the legal ramifications of mobile wireless communications that provide conve- nience, information, and entertainment to people in existing vehicles. a. Connected Vehicle DSRC Safety Systems.—Con- nected vehicle safety systems use specialized DSRC transceivers to send and receive real-time vehicle data over ad hoc V2V vehicle networks. NHTSA has announced that the agency plans to require this type of connected vehicle communications equipment in all new passenger cars and light trucks.214 USDOT initiated DSRC connected vehicle tech- nologies just after the turn of the 21st century, as part of a USDOT research program called VII (Vehicle Infrastructure Integration). In 1999, the Federal Communications Commission (FCC) had assigned 75 megahertz of spectrum at 5.850–5.925 GHz (often referred to as the 5.9 GHz spectrum) solely for vehicle safety and mobility communica- tions over DSRC.215 The initial VII concept was to provide human drivers real-time information about the infrastructure (curves, bridges, embankments), as well as what other nearby vehicles (particularly not yet visible vehicles) were doing. The VII pro- gram developed a DSRC radio communications system over the FCC-allocated radio spectrum. DSRC radio technology has the capacity to trans- mit vehicle operation data at high speeds and with low latency. DSRC communications between a vehicle and the roadside infrastructure are called V2I. DSRC communications from one vehicle to another vehicle are called V2V. DSRC communica- tions that transmit vehicle data to all sorts of mobile devices are called V2X. In all cases, DSRC’s function is wireless transmission of vehicle operational data. This data can provide useful operational inputs for driverless vehicles. In 2014, NHTSA announced a regulatory initia- tive to require DSRC “Connected Vehicle” radio transceivers as mandatory safety equipment in all new passenger vehicles and light trucks in the United States.216 This announced, but not yet for- mally proposed, mandatory V2V safety requirement is not aimed specifically, much less solely, at driver- less vehicles. According to NHTSA, required V2V operational data will be exchanged anonymously over ad hoc networks for the purpose of warning drivers of conventional vehicles.217 Still, such a DSRC requirement would also provide driverless vehicles a valuable source of real-time data about nearby vehicle operations. Four serious uncertainties cloud the potential for near-term requirements regarding DSRC con- nected vehicle V2V safety data transmissions. First, the FCC is under Congressional pressure to re-allocate parts of the now-dedicated 5.9 GHz DSRC spectrum to other types of wireless users.218 Other, non-vehicle uses of this spectrum could cause interference that would degrade the useful- ness of DSRC real-time vehicle communications to the point that that V2V Connected Vehicle commu- nications could become unreliable, particularly in congested urban areas. Interference issues con- tinue to be under study.219 Second, when NHTSA announced the agency’s intention to require connected vehicle DSRC trans- ceivers in all passenger vehicles and light trucks, 214 Federal Motor Vehicle Safety Standards: Vehicle-to- Vehicle (V2V) Communications, 79 Fed. Reg. 49,270 (pro- posed Aug. 20, 2014) (to be codified at 49 C.F.R. pt. 571) (hereinafter “NHTSA ANPRM”). See also NHTSA, supra note 207. 215 Amendment of Parts 2 and 90 of the Commission’s Rules to Allocate the 5.850–5.925 GHz Band to the Mo- bile Service for Dedicated Short Range Communications of Intelligent Transportation Services, 14 FCC Rcd. 18221 (Oct. 21, 1999). 216 NHTSA ANPRM, supra note 214. See also nAT’l highWAy TrAffiC sAfeTy Admin., U.s. deP’T of TrAnsP., vehiCle-To-vehiCle CommUniCATions: reAdiness of v2v TeChnology for APPliCATion xiii (2014) (hereinafter “NHTSA Readiness Report”) http://www.nhtsa.gov/ staticfiles/rulemaking/pdf/V2V/Readiness-of-V2V- Technology-for-Application-812014.pdf. 217 NHTSA ANPRM, supra note 214, 79 Fed. Reg. 49,270, 49,272 (“[W]e plan to propose to require that new vehicles be equipped with DSRC devices, which will enable a variety of applications that may provide various safety- critical warnings to drivers.”). 218 Revision of Part 15 of the Commission’s Rules to Permit Unlicensed National Information Infrastructure (U-NII) Devices in the 5 GHz Band, 28 FCC Rcd. 1769 (Feb. 20, 2013). See also Rebecca Blank & Lawrence E. Strickling, U.S. deP’T of CommerCe, Evaluation of the 5350–5470 MHz and 5850–5925 MHz Bands Pursuant to Section 6406(b) of the Middle Class Tax Relief and Job Creation Act of 2012 (Jan. 2013), http://www.ntia.doc.gov/ files/ntia/publications/ntia_5_ghz_report_01-25-2013.pdf. 219 Michael O’Rielly & Jessica Rosenworcel, Driving Wi-Fi Ahead: the Upper 5 GHz Band, offiCiAl fCC blog (Feb. 23, 2015), https://www.fcc.gov/blog/driving-wi-fi- ahead-upper-5-ghz-band.

25 substantial objections were raised about the con- tinuing absence of adequate measures to protect both privacy and security as well as to prevent the use of V2V for surveillance. Some of these privacy, security, and surveillance concerns are based on DSRC connected vehicle design features in which unencrypted vehicle operational data (the Basic Safety Message, or BSM) are transmitted “in the clear” from vehicle to vehicle. See Section VII, infra, for a discussion of these issues in greater detail. Third, there are legal objections to NHTSA’s announced intention to propose agency regulations that would require a DSRC transceiver to be embed- ded in every new passenger vehicle and light truck. There is no express statutory authorization for such an agency requirement.220 The need for legislative authorization further increases the possibility that NHTSA’s intended DSRC connected vehicle require- ment may be delayed or even blocked by Congress. Fourth, some transportation technology experts are beginning to view DSRC as 1990s technology that needs reassessment in light of newer and better communications technologies.221 So far, alternative communication technologies, such as those used in commercial mobile wireless applications described below, have not yet attained the speed and low latency that make DSRC essential for vehicle safety communications. Nevertheless, improvements in commercial mobile wireless technologies may pro- vide alternative ways to transmit specialized vehicle operation data wirelessly in real-time.222 If such improved wireless communications technologies can attain the functionality of DSRC, driverless vehicles may transmit and receive vehicle safety data over these wireless channels, instead of the currently planned connected vehicle DSRC safety systems. b. Connected Vehicle Wireless Mobility Applica- tions (Mobile Wireless).—Connected Vehicle Wire- less Mobility Applications (Mobile Wireless) are different from the narrowly focused, standardized DSRC Connected Vehicle Safety Systems discussed in the previous section. Mobile wireless communica- tions comprise a heterogeneous group of technolo- gies using commercial wireless networks (currently 4G and LTE, but by the time driverless vehicles are introduced, probably some form of 5G). These wire- less technologies send and receive a wide range of data, including navigation assistance, traffic, weather, phone conversations, email, and entertain- ment programming, as well as vehicle operation data. Mobile wireless services also include satellite services, such as Sirius XM Satellite Radio, that transmit digital signals into moving vehicles, under circumstances where slower transmission speeds and higher latency do not interfere with the purpose of the transmission. For short distances within a vehicle, Bluetooth is typically used for wirelessly communicating among devices. Mobile wireless technologies, such a smart- phones, can be brought into a vehicle to provide navigation and other information to enhance vehicle mobility and convenience. Applications include existing commercial services that provide navigation and parking advice, automatic acci- dent reporting, weather, and traffic reports. Some wireless services provide only audio-video enter- tainment and information. Others facilitate Internet connections. Apple and Google provide the two main vehicle platforms that allow smartphone (phone and Inter- net) functions to appear on a vehicle’s dashboard display screen and enable smartphone control by using the vehicle’s controls, including voice controls. Apple’s interface, called CarPlay, was launched in March 2014. Google’s similar interface, called Android Auto, launched in June 2014. In addition, many vehicle manufacturers embed closed-network wireless communications platforms that automatically communicate data regarding vehicle parts and operations back to the vehicle’s manufacturer. Some of these closed wireless sys- tems also carry infotainment, navigation, and auto- matic crash notification (ACN) services. Automotive operating systems that have been used to run this type of embedded vehicle connectivity include Micro- soft Embedded Automotive, open-source MeeGo, and QNX Car from Research in Motion. The most advanced vehicle-embedded communications sys- tems offer cross-platform mobile access to phone, Internet sites, infotainment, and email, as well as provide data communication between a vehicle’s automotive systems and its manufacturer. Hackers used this communications feature in a Jeep Chero- kee to tap into vehicle control systems so that they 220 Jenna Greene, Car Talk: Sharp Turns Ahead, nAT’l L.J. (Dec. 1, 2014), http://www.nationallawjournal.com/ id=1202677551065/Car-Talk-Sharp-Turns-Ahead. See also rAnd rePorT, supra note 181, at 100 (discussing limita- tions to NHTSA regulatory authority). 221 Brad Templeton, Will Robocars Use V2V at All?, ROBOHUB (Feb. 3, 2015), http://www.automotiveitnews. org/articles/share/559041/. 222 Vehicle-to-Vehicle Communication Technology Industry Awareness Driven by Driverless Cars; DSRC to be Challenged by 5G in the Next Decade, Says ABI Research, fin. mirror (Jan. 20, 2015), http://www.financialmirror. com/newsml_story.php?id=32225. Cf. Randal O’Toole, Policy Implications of Autonomous Vehicles 1, 5–8 (2014), http://object.cato.org/sites/cato.org/files/pubs/pdf/pa758_1. pdf (describing the extensive driving data most driv- ers already get from mobile wireless technology through platforms such as Google Maps, Apple Maps, and various weather mobile apps).

26 could remotely take over operational control of the vehicle while it was being driven on a highway.223 In March 2014, the Federal Highway Administra- tion (FHWA) published a Federal Register Notice requesting information about Connected Vehicle Mobility Applications “that leverage the full poten- tial of trusted communications among connected vehicles, travelers, and infrastructure to better inform travelers, enhance current operational prac- tices, and transform surface transportation systems management.”224 This research program seeks “applications that synergistically capture and uti- lize new forms of connected vehicle and mobile device data to improve multimodal surface trans- portation system performance and enable enhanced performance-based systems management.”225 FHWA apparently seeks to leverage connected vehicle data for use in commercial applications, as well as traffic management and safety programs. In 2015, USDOT launched a connected vehicle research program to encourage “Dynamic Mobility Applications.” This USDOT program will “combine connected vehicle and mobile device technologies in innovative and cost-effective ways to improve trav- eler mobility and system productivity, while reduc- ing environmental impacts and enhancing safety.”226 The Dynamic Mobility Applications program envi- sions competitive commercial development, with the federal government playing “an appropriate and influential role as a technology steward for the con- tinually evolving integrated transportation [infor- mation] system.”227 The program seeks development of ways to use connected vehicle data for traveler convenience, safety, environmental, and transporta- tion management functions. USDOT encourages, but does not generally regu- late, Mobile Wireless applications, except insofar as they may pose safety hazards in the form of driver distractions. In 2013, NHTSA published voluntary guidelines that restrict visual and tactile access to many types of in-vehicle wireless devices and dis- plays, such as those used in Mobile Wireless applica- tions.228 Because these guidelines only affect driver- facing interfaces, they will not apply to driverless vehicles. Since automated, semi-autonomous, or partly self-driving vehicles all have drivers subject to distraction, the driver distraction guidelines would apply. So far, NHTSA has brought no formal enforcement actions related to Mobile Wireless applications that have allegedly distracted drivers of conventional cars. Nevertheless, NHTSA has warned that the agency may initiate enforcement if in-vehicle electronic devices contain safety-related defects or cause driver distraction when human drivers are in control of vehicles. 3. FCC Communications Regulation The FCC allocates the communications spectrum and licenses both telecommunications devices and wireless telecommunications carriers that transmit communications to and from vehicles. In addition, the FCC’s E911 regulations, adopted in 2010, require wireless mobile phone communications to provide location information (primarily from GPS). Although there have been suggestions that the FCC adopt specific licensing regulations with regard to providers of vehicle-based mobile wireless ser- vices (which the FCC generally refers to as telemat- ics), the FCC has so far only asserted general juris- diction over wireless communications devices and wireless service providers. Aside from licensing the 5.9 GHz spectrum (5.850–5.925 GHz) for use by DSRC vehicle safety and mobility services, the FCC does not yet specifically regulate connected vehicle communications platforms. Further regulation of vehicle communications systems by the FCC is pos- sible, as communications from connected vehicles become more widespread.229 4. Vehicle Connection Security Issues Among the most serious challenges faced by con- nected vehicles, whether human-driven or driver- less, are heightened cybersecurity threats. In the context of Mobile Wireless applications, security threats can be difficult to guard against because there are so many sources both of data input and of types of communications. In a multi-connection wireless communications setting, identifying, iso- lating, and preventing security threats from hack- ers, malware, defective equipment, and other cybersecurity threats is especially difficult. See Section VII, infra. According to a recent report, “A new car may have more than 145 actuators and 75 sensors, which pro- duce more than 25 GB of data per hour. The data is analyzed by more than 70 onboard computers to 228 Visual-Manual NHTSA Driver Distraction Guide- lines for In-Vehicle Electronic Devices, 78 Fed. Reg. 24818 (Apr. 26, 2013). 229 See the discussion of potential FCC privacy regula- tion, infra at notes 508–511. 223 miller & vAlAsek, supra note 203, at 20–33. The system hacked was a QNX-based system called Uconnect. 224 Connected Vehicle Pilot Deployment Program; Request for Information, 79 Fed. Reg. 14105, 14105 (Mar. 12, 2014). 225 Id. 226 Connected Vehicles CV Pilots Deployment Project, U.S. deP’T of TrAnsP. (Nov. 7, 2014), http://www.its.dot. gov/pilots/cv_pilot_progress.htm. 227 Id.

27 ensure safe and comfortable travel.”230 Connected vehicle mobility applications access this vehicle data to provide feedback data to the manufacturer of the vehicle. They also offer attractive hacker targets. Mobility applications often include “infotainment systems, engine management units, and onboard diagnostic units, radios operating at different fre- quencies, GPS receivers, transponders, Bluetooth devices, and cell phone chips.” As a result, “malware in any subsystem could compromise the safety of not only the people in the car, but also those around them.”231 Research is underway with regard to potential security threats in the context of connected vehicle communications systems. Development of security solutions for connected vehicle communica- tions is discussed in Section VII, infra. D. Manufacture and Sales It is unlikely that automobile original equipment manufacturers (OEMs) will build everything that goes into driverless vehicles, including all parts and systems. Instead, driverless vehicle manufacturers will, almost certainly, integrate parts and technolo- gies from component manufacturers into driverless vehicles.232 Specialized companies, such as Bosch, Continental, and many other vehicle parts suppli- ers, are developing driverless and automated vehicle parts and modules.233 Technology companies, such as Google, Inc., also develop components or modules for assembly into the company’s driverless vehicles. Because of the need to integrate multiple automated systems within a driverless vehicle, it is most likely that driverless cars and trucks will be manufactured solely as original equipment, rather than as after- market driverless vehicle retrofit modules or kits.234 Most conventional cars and trucks are currently sold through intermediaries, known as dealers. It is possible that some driverless cars will be sold directly by automobile companies, rather than through deal- ers. For example, Tesla Motors currently markets its advanced electric cars directly to purchasers, with- out an intermediary dealer. The vehicle manufac- turer handles continuing warranty service and maintenance. In driverless vehicles, software, such as mapping, is likely to require continuing mainte- nance and more frequent updating than physical aspects of the vehicle. Frequent software and firm- ware updates for driverless vehicles are expected to be wirelessly downloaded from manufacturers. Continuing need for updates, mapping, and other programming modifications may tether a driverless vehicle to its manufacturer throughout the life of the vehicle. As discussed elsewhere in this report, the need for ongoing changes in driverless vehicle programming and systems are likely to affect prod- ucts liability for harm resulting from driverless vehicle programming modifications. Virtually con- tinuous vehicle information exchanges with manu- facturers through mobile wireless communications is already a common feature of some advanced motor vehicles, particularly electric vehicles.235 E. Vehicles Not Considered Driverless Vehicles Before the driverless vehicles discussed in this report become available, several forms of highly automated vehicles will likely be in use. As dis- cussed earlier, part-time or partially self-driving vehicles are not, for the purposes of this report, con- sidered driverless vehicles. 1. Vehicle Platoons Vehicle platooning has been a subject of enthusi- astic discussions for a long time. Applications of vehicle platoons, from truck convoys to “car-trains,” rely on wireless communications connecting one platooned vehicle with the next, and the rest of the chain. These vehicles currently require human driv- ers to manage entry into and exit from the platoon, although platooned vehicles are not under active driver control during the part of the journey when they are attached to the platoon. For the purposes of this report, platooning is an example of a highly promising connected vehicle technology that pro- vides opportunities for a degree of human driver passivity. However, a vehicle in a platoon is not a driverless vehicle, in the sense that a human driver 230 Max Glaskin, Safe and Secure, vision zero inT’l, June 2014, at 40. 231 Id. 232 Gabe Nelson, Google in talks with OEMs, suppli- ers to build self-driving cars, AUTomoTive neWs (Jan. 14, 2015), http://www.autonews.com/article/20150114/ OEM09/150119815/google-in-talks-with-oems-suppliers- to-build-self-driving-cars. 233 Joseph Szczesny, Continental Ups the Autonomous Car Ante, The deTroiT bUreAU (Aug. 18, 2014), http://www. thedetroitbureau.com/2014/08/continental-ups-the-au- tonomous-car-ante/. Stephen Edelstein, Bosch expects to see self-driving cars in 10 years, digiTAl Trends (Jan. 20, 2015), http://www.digitaltrends.com/cars/bosch-expects- see-self-driving-cars-10-years/. 234 Installation of driverless vehicle components is prob- ably too difficult to make aftermarket versions of driver- less vehicles realistic. Michigan law exempts from liability for injuries from product defects not present at the time of manufacture both manufacturers and subcomponent manufacturers if vehicles are converted into automated motor vehicles. miCh. ComP. lAWs § 600.2949b (2014). 235 See Francesca Svarcas, Turning a New LEAF: A Privacy Analysis of CARWINGS Electric Vehicle Data Collection and Transmission, 29 sAnTA ClArA ComP. & high TeCh. L.J. 165 (2012).

28 is entirely superfluous. In the future, driverless vehicles may be able automatically to attach and detach from vehicle platoons. But that does not appear to be a near-term option. Early applications of this technology have been developed as truck platoon systems such as that provided by Peloton Technology in the United States.236 In Europe, the European Commission established the SARTRE Project (Safe Road Trains for the Environment). Successfully developed SARTRE vehicle platoons operated on normal pub- lic highways and demonstrated significant environ- mental, safety, and comfort benefits, mostly in the context of long-haul trucking.237 Since a follow-the-lead–truck approach still retains drivers, it does not involve vehicles that are, in pres- ent versions, driverless. A driver operates the lead truck; in the following trucks, drivers remain in the vehicles, although they do not exercise active vehicle control while they are in the platoon. In the United States, legal impediments to lawful operation of truck platoons take the form of state laws that specifically ban “truck convoys.”238 Other state statutes set specific minimum spaces between vehicles under “following too close” prohibitions.239 2. Remotely Controlled Vehicles Remote control over a vehicle by an external oper- ator does not make the vehicle driverless. Although no human driver may be present in the vehicle, a remotely controlled vehicle does not control its own operation. Control by external operators simply moves the vehicle’s “driver” from being a human inside the vehicle to someone outside the vehicle. Remotely controlled vehicles are often associated with familiar childhood toys. In commercial ver- sions, they are used in mining and military opera- tions, often in the form of very large-scale trucks, digging equipment, and UGVs (unmanned ground vehicles).240 In the context of rail transport, remote control of trains has been a controversial feature of railroad operation for a long time, partly because of job displacement of human train operators.241 This report instead focuses on driverless vehicles whose operations are not controlled by any human driver either inside or outside the vehicle. F. Driverless Vehicle Uses and Deployment Slightly different legal rules and policies will apply to different ways in which driverless vehicles may be used. For example, driverless trucks in inter- state commerce likely will be subject to federal regu- lation with regard to minimum insurance require- ments; driverless passenger cars will instead be subject to state insurance regulation. Potential uses for driverless vehicles include: • Individually owned personal/family transpor- tation; • On-demand personal-mobility services in urban areas; • Rental vehicles for short-term mobility and transport needs; • Long-haul movement of goods and commodi- ties; • Commercial local delivery services; • Paratransit driverless vehicles (services for persons with disabilities); • Fleets owned by corporations or other entities; • Fleet ownership by groups of users for coopera- tive use; and • Urban low speed vehicles on limited roadways. 236 PeloTon, http://www.peloton-tech.com/about/ (last visited Sept. 20, 2015). Peloton is an automated vehicle technology company that utilizes vehicle-to-vehicle com- munications and radar-based active braking systems, combined with sophisticated vehicle control algorithms, to link pairs of heavy trucks. The safety systems are always active, and when the trucks are out on the open road, they can form close-formation platoons. 237 sArTre ProJeCT, http://www.sartre-project.eu/en/ Sidor/default.aspx (last visited Sept. 20, 2015). See also Ian Norwell, Road Trains on Track?, TrAnsP. engineer (Apr. 28, 2014), http://www.transportengineer.org.uk/ transport-engineer-features/road-trains-on-track/60995. 238 E.g., CAl veh. Code § 21705 (West 2000) (“Motor vehicles being driven outside of a business or residence district in a caravan or motorcade, whether or not towing other vehicles, shall be so operated as to allow sufficient space and in no event less than 100 feet between each vehicle or combination of vehicles so as to enable any other vehicle to overtake or pass.”). 239 E.g., N.Y. veh. & TrAf. lAW § 1129 (McKinney 2011): Following too closely. (c) Motor vehicles being driven upon any roadway outside of a business or residence district in a caravan or motorcade whether or not tow- ing other vehicles shall be so operated as to allow suf- ficient space between each such vehicle or combination of vehicles so as to enable any other vehicle to enter and occupy such space without danger. 240 See, e.g., Mark L. Swinson, Robotics, Military, in 15 enCyCloPediA of miCroComPUTers 99, 105–07 (Allen Kent & James G. Williams eds., 1995); Horst Wagner, Mining Technology for Surface and Underground Mining—Evolv- ing Trends, in 1 mining in The 21sT CenTUry: qUo vAdis? 35, 47–48 (A.K. Ghose & L.K. Bose eds., 2003); Press Re- lease, Rio Tinto Improves Productivity Through the World’s Largest Fleet of Owned and Operated Autono- mous Trucks (June 9, 2014), http://www.riotinto.com/ media/media-releases-237_10603.aspx. The mining trucks Rio Tinto claims to be “autonomous” are in fact operated by remote control. Id. 241 fed. rAilroAd AdminisTrATion, U.s. deP’T of TrAnsP., remoTe ConTrol loComoTive oPerATions: resUlTs of foCUs groUPs WiTh remoTe ConTrol oPerATors in The UniTed sTATes And CAnAdA 8, DOT/FRA/ORD-06/08 (2006).

29 Which of these driverless vehicle uses will develop earliest and which applications would provide the most demand for driverless vehicles is difficult to predict. However, it is possible to sort out some of the factors and circumstances likely to encourage use of driverless vehicles, as well as some factors that would tend to discourage use of driverless vehicles. 1. Factors Encouraging Driverless Vehicle Use Factors that are likely to encourage driverless vehicle market interest or to stimulate purchase of driverless vehicles will include enhanced safety, con- venience, and efficiencies such as the ability of rid- ers to perform other tasks or to rest. Availability of fleets of driverless vehicles for on- demand use appears to be attractive to many poten- tial users, particularly in urban areas. Repeated journeys along the same roads (commuting to work or to school), frequent slow-moving traffic or traffic stoppages, opportunities to multitask or to do noth- ing, as well as individual personal preferences for solitary personal mobility without a human driver are all likely to encourage interest in purchasing or using a driverless vehicle. These factors tend to be present primarily in urban and suburban settings. More intense use of urban area roadways will tend to increase the accu- racy of real-time roadway maps and to provide more data sources regarding traffic and weather condi- tions. Special road markings, or beacons, automati- cally transmitting data to driverless vehicles would also be more economically justifiable in higher-use urbanized areas. Overall, economies of scale suggest more densely traveled areas as more conducive for use by driverless vehicles. Because driverless vehicles will generate quite a number of public goods such as environmental and public safety benefits,242 tax or regulatory incentives may provide an additional factor encouraging pur- chase of driverless vehicles. 2. Factors Discouraging Driverless Vehicle Use Factors likely to discourage interest in driverless vehicles include cost; psychological queasiness about loss of control; roadway risks from other vehicles and the infrastructure, as well as from non-vehicle road users (pedestrians, bicycles, etc.); concerns about surveillance and tracking of individuals; and insecurity about potential defects in and hacker attacks on driverless vehicle technical systems. It is unclear whether driverless vehicles will be preferred for long or short journeys. Initially, driverless vehi- cles may be rarely used until controlled operating environments, such as segregated roadways for driverless vehicles, are established. At the outset, increased cost will be an impor- tant factor discouraging many would-be purchas- ers of driverless vehicles. According to Morgan Stanley analyst Ravi Shanker, at today’s prices, full driverless capability is estimated to add about $10,000 to the cost of a car.”243 The Boston Consult- ing Group estimates that in 2025 a driverless vehi- cle will add $9,800 to the vehicle’s base price.244 Increased production of driverless vehicles should bring costs down. Corporate or cooperative purchases of driverless vehicles for on-demand use will face uncertainties about demand patterns, as well as high front-end capital costs. 3. Specialized Driverless Vehicle Environments In the transition to widespread use of driver- less vehicles, these vehicles will have to be able to cope with human-driven vehicles. As noted earlier, at least initially, driverless vehicles may first be used in special controlled environments, such as areas restricted to low-speed vehicles or restricted travel lanes. a. Urban Personal Mobility On-Demand Services. —When people are asked about the type of driver- less vehicles they want to be available first, they usually point to on-demand personal mobility ser- vices for short trips.245 On-demand driverless car services promise convenience and privacy in trans- porting people to and from local destinations in urban areas where population density makes such transport-on-demand profitable. Existing online ride services, which are sometimes called “ride-shar- ing,” “ride-hailing,” or “Transportation Network Companies” (TNCs), are smartphone applications popularized by Uber, Lyft, Sidecar, and similar ven- tures. They are a frequently mentioned business model for use of driverless vehicles. Summoning a vehicle without a driver seems to be both potentially 242 RAND rePorT, supra note 181, at 9. 243 Morgan Stanley, Autonomous Cars: The Future is Now (Jan. 23, 2015), http://www.morganstanley.com/ articles/autonomous-cars-the-future-is-now/. 244 bosTon ConsUlTing groUP rePorT, supra note 188, at 15. 245 See, e.g., Antonio Loro, Driverless Taxis: The Next Next Big Thing in Urban Transportation?, PLANETIZEN (May 6, 2014), http://www.planetizen.com/node/68657. See also Bora Alp Baydere, Kelechi Erondu, Daniel Espinel, Siddarth Jain & Charlie Ritter Madden, Car- Sharing Service Using Autonomous Automobiles 3 (Spring 2014) (unpublished manuscript), http://web.stanford.edu/ class/me302/PreviousTerms/2014-06Car-SharingService UsingAutonomousAutomobiles%28paper%29.pdf.