Collision Avoidance of a Kinodynamically Constrained System from Passive Agents

Robot motion planning in dynamic environments is significantly difficult, especially when the future trajectories of dynamic obstacles are only predictable over a short time interval and can change frequently. Moreover, a robot’s kinodynamic constraints make the task more challenging. This paper proposes a novel collision avoidance scheme for navigating a kinodynamically constrained robot among multiple passive agents with partially predictable behavior. For this purpose, this paper presents a new approach that maps collision avoidance and kinodynamic constraints on robot motion as geometrical bounds of its control space. This was achieved by extending the concept of nonlinear velocity obstacles to incorporate the robot’s kinodynamic constraints. The proposed concept of bounded control space was used to design a collision avoidance strategy for a car-like robot by employing a predict-plan-act framework. The results of simulated experiments demonstrate the effectiveness of the proposed algorithm when compared to existing velocity obstacle based approaches