and computational limitations inherent in the MAST robot vision. All research tasks in the MAST CTA should directly address these issues.
Possibly missing from this research portfolio, especially for the MAST CTA, are power-aware computing, intelligent power management, and algorithms for managing bandwidth. In addition, some areas of important teaming research were not mentioned, such as the overall organization of the teaming (e.g., centralized versus decentralized intelligence across teams, hierarchical teaming, or other organizational strategies) and multi-robot decision making at the mission level.
Many individual projects address the issue of robot mapping; however, these project teams are not talking with one another, which raises the following questions: What mission objectives are being addressed by these different mapping projects? Why not perform comparisons across mapping projects for a variety of Army applications? Why not study whether the key strengths of different mapping algorithms can be merged into a single approach? Why not study the tradeoffs for each type of mapping for various Army-relevant applications? For example, multiple projects are under way to autonomously map out the interior of buildings, and these projects appear to be making solid progress. Because most buildings worldwide are of multiple stories, it is time to conduct a down-select of interior mapping systems and focus the resulting system on some stair-climbing approach such as was demonstrated so effectively in the autonomous stairwell assent project. It is very important to consider all stair-climbing concepts, such as tracked, legged, and wheeled platforms, in the overall program of autonomous mapping of the interior of multiple-storied constructions. The importance of instantaneous interior building mapping to the military mission remains somewhat dubious. How important would it be to map the inside of the building if it could be instantaneously and reliably determined that there were no combatants or other dangers inside it?
Overall Technical Quality The robotics enterprise is grappling with difficult problems that have been around for a long time. The enterprise’s approach is to pursue many different algorithms for these problems, and then at some point in the future decide which is most beneficial for a given Army application. Research advances are being made to the state of the art, mostly of an incremental nature. Not all of the research approaches are well-justified (e.g., the hybrid mapping research, which pursues techniques based on the available sensor, rather than best approach to solving the problem). A better justification of the research approaches being pursued is needed in many areas of the intelligence research—particularly in terms of military requirements. Robot research in the civilian market continues to dwarf ARL’s efforts. ARL should strive to fill the military-relevant gaps left in the civilian research, not duplicate that research.
Multi-Robot Persistent Surveillance Planning as a Vehicle Routing Problem
This work addresses a long-existing problem using a new technique, that of multi-vehicle patrolling. The objective is to solve the problem exactly. Comparing the approach to a baseline of the traveling salesman solution is commendable. The ability to generate solutions that vary the visit times is a nice side-effect, because it reduces the external predictability of the system. This research should be presented more carefully, recognizing that the approach is actually not generating a globally optimal result and that heuristic approaches can have value over exact solutions, specifically for solving large-sized problems in this probably intractable domain. Using this technique as a foundation for addressing the general task allocation problem with spatio-temporal constraints is an idea worth pursuing. Feedback should be added to the approach in order to close the loop on the control.
There is a lack of knowledge regarding similar past and recent relevant activities conducted by NASA, DARPA, and other government organizations. Additionally, this work should better address the