Navigation is a repetitive cycle of sensing, planning and action.

Red Team robots estimate position and state by evaluating GPS, inertial measurement and odometry data (Applanix). The vehicles perceive terrain with laser sensors, cameras (SAIC) and radar (Duke, Boeing). A laser scanner and cameras are mounted on a stablizing three-axis gimbal which directs sensors into terrain regions incompletely scanned by fixed sensors. Sensor data is fused into a composite cost model of the terrain in which less navigable areas are assigned higher cost values.

Path candidates are generated and evaluated as s-curved paths with regard to the pre-planned route, the cost model and vehicle dynamics. The curves blend into a smooth line as path tracking commands the vehicle to follow the planned path. Race logic software (Agent-Oriented Software) layers over navigation to set pace and react to contingencies.

Animation & Video

Sensor Data and Gimbal Pointing Green fields cover terrain scanned by left and right laser sensors, the red field covers terrain scanned by a central laser sensor. Two purple cubes bound the area at which the gimbal aims its laser line scanner and stereo camera pair to fill holes in the data.	S-curves The green X farthest from the vehicle marks the next position the vehicle will attempt to gain. Curves represent paths to the target position. Areas marked in green are perceived as negotiable terrain, red is perceived as non-negotiable.	Path Localization Path traversal costs based on laser sensor data are displayed in blue with lighter shades representing safer routes. The orange line is the vehicle's pre-planned path, the green line is the localized path.
Gimbal Motion	Driving	Radar Page