Obstacle avoidance and speed control in a mobile vehicle equipped with a compound eye

摘要

Shows that the use of onboard visual motion sensors has advantages in visually guided navigation. An agent equipped with a given array of motion detectors detects obstacles all the more distant when it is running faster. Its "radius of vision" becomes simply proportional to speed, with the interesting consequence that the agent can easily adapt its speed to the environmental context. The simulation study presented shows that the newly designed visuomotor system allows the robot to reach a target while avoiding obstacles in cluttered environments, adapt its own speed to the situation, follow walls, retreat from dead-ends and bypass impenetrable clusters of obstacles. These performances result all from one and the same visuomotor algorithm and do not require any "arbitration" between different "behaviors" as is the case with the subsumption architecture. Nor do they require any kind of high level reasoning, world model and long term memory. The described system could be part of a hierarchical system and carry out the basic and reputedly difficult task of fast piloting in uncertain environments, for example, between intermediary waypoints that would be specified by a high level mission planner.