Obstacle detection and avoidance with noisy measurements using danger zone concepts

Potential collision danger increases if a vehicle gradually moves closer to other vehicles or obstacles. If the collision time can be predicted before the vehicle is dangerously close to other vehicles or obstacles, the vehicle can devise an effective evasive maneuver or reroute its path, thus avoiding collision. This paper proposes an obstacle avoidance algorithm for unmanned vehicles in an unknown environment. The proposed avoidance algorithm uses the danger zone concept to determine whether the obstacle or surrounding vehicle will cause a possible collision. The danger zone is constructed around the vehicle by using semi-algebraic functions. The possible collision can be predicted by checking the sign of these semi-algebraic functions. The semi-algebraic functions for constructing the danger zone apply the relative velocities between the vehicle, and other vehicles or obstacles and can be used for both two- and three-dimensional cases. Several avoidance schemes are provided, with examples to demonstrate the concept of the proposed approach.