Bi-Risk-RRT Based Efficient Motion Planning for Autonomous Ground Vehicles

Autonomous ground vehicles (AGVs) have been deployed in various working environments. Human-AGV coexisting environments introduce many challenges into the motion planning procedure. The planner should be smart enough to guide the AGV to move safely and smoothly. Besides, the efficiency of the planner is essential for real-time response. Risk-based motion planning algorithms consider the risk of collision with dynamic and static obstacles to deal with planning problems in human-AGV coexisting environments. However, their performance needs to be further improved. The bidirectional search sampling strategy is widely used to obtain real-time response capability. However, when considering kinodynamic constraints, the bidirectional search sampling strategy needs to solve the two-point boundary value problem (TBVP), which is not always solvable for all kinds of kinodynamic constraints. To achieve efficient AGV navigation in the environment with pedestrians, we propose a risk-based algorithm with a bidirectional search sampling strategy, namely bidirectional Risk-RRT (Bi-Risk-RRT). Additionally, by using the other direction as a heuristic, Bi-Risk-RRT bypasses TBVP and retains the advantages of the bidirectional search sampling strategy. The simulation experiments are conducted in various challenging environments, and the results demonstrate the efficiency of the proposed algorithm.