2003-2004 Assessment of the Army Research Laboratory (2005)

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.

9 Robotics INTRODUCTION Robotics research and development R&D at ARL is focused on semiautonomous navigation for small and medium-sized vehicles. The R&D has four components: (1) Perception refers to the mecha- nism by which a robot senses and perceives the environment. (2) The control function takes perpetual features and translates them into actuator actions (e.g., acceleration, braking, and turning) that control vehicle navigation. (3) Robot supervision is an aspect of a semiautonomous vehicle system that allows a human operator to intervene and take control of the vehicle if required. (4) Mission packages refer to end-user applications for which autonomous robots are being developed. As presented to the Robotics Review Team, the focus of ARL’s robotics activities is the development of semiautonomous ground vehicle technologies to support operations and provide situation awareness capabilities in the battlefield. One of the goals of the robotics activities is to achieve control of many robots by one or a few humans. That goal is addressed by human workload measurement, important to ARL experiments. The Board believes that this goal will not be achieved soon. As soon as mission requirements are considered, the personnel needed to manage each robot’s activities escalate. For a reconnaissance mission, for example, the sensory output is intended for the benefit of human observers, and the routing of the information obtained by the robot must be managed interactively. For targeting, it will be a long time before human verification is not required. Perhaps a more feasible near-term goal is single-operator management of several Multifunction Utility/Logistics Equipment (MULE) robots. The Robotics Review Team, assembled at the request of the ARL Director, met in May 2004 at the Army War College to conduct a crosscutting technical assessment of programs within the Army Re- search Laboratory (ARL) relating to ongoing efforts in the field of robotics. That was the first year that such a review of the ARL robotics program had been conducted. The review covered activities from multiple ARL directorates. The reviewers focused on the overall 51 

52 2003–2004 ASSESSMENT OF THE ARMY RESEARCH LABORATORY quality of the efforts and on integration of those efforts among themselves and with other ARL objec- tives. The review included presentations of technical projects, a live demonstration of field robotics, and a significant dialogue between the reviewers and presenters. The Review Team members, drawn from existing panels of the Army Research Laboratory Technical Assessment Board, represented a composite of academic and industrial expertise. The review was hosted by the Weapons and Materials Research Directorate (WMRD), which also manages the Robotics Collaborative Technology Alliance (CTA). The Review Team was briefed on a broad array of ARL robotics activities. The briefings were organized into six components: technology demonstrations, program overview, research in perception, research in intelligent control, research in human-robot interfaces, and research pertaining to mission packages. The ARL in-house presenters represented more than 60 ARL staff members and several contractors and collaborators from the CTA. The ARL robotics projects presented to the panel are supported by more than $20 million per year of research funding. ACCOMPLISHMENTS AND OPPORTUNITIES Most Significant Advances The Robotics Review Team considered the technical quality of the ARL core robotics research to be high. Indeed, much of it is competitive with state-of-the-art, world-class research in applied robotics. ARL has produced three notable ground vehicle robot platforms: the Soldier Unmanned Ground Vehicle (SUGV), the Multifunction Utility/Logistics Equipment (MULE), and the Armed Robotic Vehicle (ARV). ARL developed the Experimental Unmanned Vehicle (XUV) as a testbed to assist in the development of autonomous navigation technology. The SUGV is a small portable robot vehicle de- signed to provide situation awareness data to the soldier. The MULE, the “soldier’s pickup truck,” is a larger semiautonomous vehicle designed to follow (in some cases precede) and support the dismounted soldier at low speeds. The ARV is a larger (approximately 9 tons), smarter, and faster robot vehicle designed with more advanced navigation and perception capabilities. Within the Future Combat Sys- tem, the MULE and ARV share the same autonomous navigation system, toward which much of the ARL robotics research has been directed. The Review Team was given a demonstration of the XUV prototype at ARL’s Fort Indiantown Gap Robotics facility in Pennsylvania. This facility provides a realistic testbed that facilitates validation and integration of algorithms for autonomous and semiautonomous vehicles. The XUV testbed allows alternative algorithms to be compared on a common platform and vulnerabilities to be investigated through realistic experiments. The Review Team appreciated the unusual complexity of the test course, which consists of a mix of road and wooded terrain. It was obvious that the development of the XUV prototype has benefited from a combination of careful engineering, teamwork, and system integration. The Review Team was positively impressed by the navigation capabilities demonstrated by the XUV on the difficult test terrain. Particularly notable were the maturity of the technology for an experimental program, the extent to which the XUV has been tested to date, and the excellent capability to recover autonomously from close encounters and software restarts. However, a more accurate percep- tion of ditches and extended laser sensing capabilities will be required to operate reliably at higher speeds and on varying terrain. Some architectural issues regarding the coupling of perception to motor skill and the coupling of perception to low-level mobility are areas of improvement worthy of future investigation. It was clear from the review that the progress of the XUV program benefited from close interaction 

ROBOTICS 53 between researchers from the Sensors and Electron Devices Directorate (SEDD), WMRD, the Human Research and Engineering Directorate (HRED), and the Computational and Information Sciences Direc- torate (CISD). Furthermore, the review revealed several new opportunities for collaboration across ARL, CTA, and Army Research Office (ARO) programs. Therefore, ARL seems well poised for development of the next generation of robot vehicle systems with more advanced sensing capabilities, higher navigation speeds, and appropriate levels and forms of effective, efficient human intervention in support of mission requirements. Opportunities and Challenges It was widely recognized by the Review Team and by ARL robotics researchers that the three principal challenges to practical and reliable autonomous vehicle deployment in the field are robustness/ reliability, navigation speed, and perception of the environment. The Review Team and the Board understand that ARL’s shorter-term objective of demonstrating the feasibility of autonomous robot vehicles precluded investigating a large number of design architectures. However, scientific progress usually requires an exploration of alternative, potentially superior architectures. The robotics activities at ARL would benefit from more attention to the science, especially in the areas of perception and intelligent control. For example, sticking to a single architecture (e.g. D* planning and Four-Dimen- sional Real-time Control System [4D/RCS] intelligent control) is acceptable for an early-prototype XUV, but other architectures should be considered for the next-generation robot vehicle. Human-robot interfaces (HRIs) is another area in which a more systematic, top-down design pro- cess would be of value. For example, while the demonstrated Operator Control Unit (OCU) is intuitively appealing, additional work remains to be done in user validation, validation of the heuristic rules used for interaction, and more sophisticated path planning. Furthermore, while a comprehensive Technology Readiness Level 6 experiment and human-factors evaluation of human workload were performed, future generations of the OCU could benefit from the inclusion of more explicit test hypotheses, more rigor in metric and attribute definition, better parameter selection, and specific exit criteria. The CTA supports research in robust local and global planning, maneuvering in dynamic environ- ments, tactical behaviors, and collaborative operations. These methods did not seem sufficiently mature to be integrated into the demonstration XUV prototype, however. The Review Team and the Board have some concern that the decoupling of the research in perception and intelligent control functions and the modularization of these functions in the architecture have created some difficulties in achieving robust navigation performance. This problem could be a major obstacle to the development of an operational system having sufficient performance and reliability for deployment in the battlefield. The obstacle described above can possibly be overcome by adopting a more basic, top-down design approach, including the development of a clearer understanding of current capabilities of technology, principal challenges to the technology, and viable strategies for meeting these challenges. This type of strategic thinking would be of benefit to research planning and would also help identify potentially useful areas for collaboration. Such open-ended thought experiments would be an excellent way to involve the problem-solving skills of academic CTA partners. Closer coupling of the CTA would likely have significant impact, perhaps even exceeding previous successes in this area (e.g., the transition of laser detection and ranging [LADAR] sensing technology to the XUV prototype). Longer-term benefits may accrue by exploiting synergies between ARL robotics activities and the ARO 6.1 Intelligent Control program. In particular, the Review Team considered groundbreaking and relevant the adaptive distributed-control research supported by the ARO and presented by the respective 

54 2003–2004 ASSESSMENT OF THE ARMY RESEARCH LABORATORY ARO program manager. The Board encourages continued and expanded interactions between ARL and ARO in areas relevant to robotics. CONTRIBUTIONS TO ARMY NEEDS AND THE BROADER COMMUNITY Contributions to Army Needs ARL’s core robotics activities are highly relevant, indeed critical, to the Army mission. ARL robotics researchers are actively deploying new perception, control, robot supervision, and mission packages on unmanned ground vehicle platforms with the potential for transition to the Future Combat System and other Department of Defense programs. The progress of the MULE and XUV prototypes has been impressive. For example, the XUV received a positive report from the Tank-Automotive Research, Development and Engineering Center at Demo III in 2003. Contributions to the Broader Community The XUV development program has strong linkages to the academic community through the robot- ics and perception projects within the CTA. The XUV platform is a state-of-the-art testbed that has been used to test and validate algorithms developed by CTA partners. This allows academic researchers to evaluate performance and failure modes of their algorithms in a realistic operational environment. RELEVANCE OF CROSSCUTTING ISSUES The robotics effort at ARL should involve more interaction between the HRED, SEDD, ARO’s 6.1 Intelligent Control program, and the Advanced Decision Architectures and the Robotics CTAs in order to improve understanding of and the interaction between physics models, feature representations, con- trol algorithms, and human factors. In particular, HRED should be more involved in assessment during the early stages of the design cycle. Such HRED involvement would be to ensure that the capabilities and limitations of the soldier are taken into account and to ensure that robotics developments are undertaken with cognizance of the systems requirements that encompass both robots and soldiers. It would be most effective for HRED to coordinate human-factors studies with the customer (e.g., at BattleLab at Fort Leonard Wood, Missouri) once a prototype is ready for customer evaluation.