Software that enables autonomous navigation for unmanned ground vehicles (UGV) is the goal of a U.S. Army research project.
Current modeling is too simplistic and fails to consider all the variables that a UGV may encounter, according to the Army small business technology transfer proposal. Current modeling systems are not able to adequately handle all of the variables vehicles can encounter, especially in rough terrain, according to the proposal. The Army needs algorithms that can take into account the characteristics of the terrain and the specific vehicle under manual control, semi-autonomous and fully autonomous operations.
"For example, in manual vehicle teleoperation scenarios, the planning algorithm could run as a background process, and safe (or unsafe) routes could be presented as visual overlays on an operator control unit, as a driver aide," according to the SBIR. "In semi-autonomous vehicle teleoperation scenarios, the planning algorithm could again be run as a background process, and safe/unsafe regions in the environment could be mapped to assistive controls applied to the vehicle, to help the operator safely guide the vehicle. In fully autonomous scenarios, the planning algorithm could form the basis of an autonomy kernel that robustly reasons about terrain difficulty and uncertainty."