Evolving (semi)-autonomous vehicles
Brian J. Capozzi, Juris Vagners · AIAA Guidance, Navigation, and Control Conference and Exhibit · 2001
Autonomous behavior implies the ability of a fleet of vehicles to adapt to changes in surroundings, mission goals, or internal system state. Adaptation includes manipulation of priorities to maintain performance levels in the presence of uncertainty. Vehicle adaptation must also be responsive to human directives - hence, (semi) autonomous behavior. We propose a framework for evolving semi autonomous vehicle behavior based on simulated evolution which uses survival of the fittest to reach a natural balance between overall mission objectives and the influence of the environment. This framework is applied to the path planning segment of the general system management problem. The resulting architecture includes the system goals directly into the search while providing a natural framework for interfacing with human management. Through simple mathematical interpretation of natural language goals, this framework provides optimalizing solutions to complex problems. This agility is demonstrated through simulation of a single vehicle adapting to changing environments, moving obstacles and objective re-prioritization. Extension of the framework to multiple vehicle systems is also explored.