A High Fidelity Multi-Sensor Scene Understanding System for Autonomous Navigation

In order for an autonomous military robot to “appropriately” navigate through a complex environment, it must have an in-depth understanding of the immediate surroundings. In the military sense, appropriate navigation implies the robot will avoid collision or contact with hazards, will not be falsely re-routed around traversible terrain due to false hazard detections, and will exploit the terrain to maximize its concealment. Appropriate autonomous navigation requires the ability to detect and localize critical features in the environment in order to respond appropriately to them. We have developed a scene understanding system based on a multi-sensor system that uses an “operator-trained” rulebase to analyze the pixel level attributes across the set of diverse phenomenology imaging sensors. Each pixel is registered to range information so we not only know what but where features are in the environment. This three dimensional labeled world model can then be used to control the speed and steering of the vehicle in an appropriate manner. In this paper we discuss our multi-sensor system, the operator trained analysis algorithm called ONAV (Opportunistic NAVigation), and the reactive control algorithm used to control the speed and steering of the vehicle.