An Approach for Multi-Objective Obstacle Avoidance Using Dynamic Occupancy Grid Map

Path planning in dynamic environments is an important guarantee for mobile robots to complete navigation tasks safely. However, human motion is unpredictable, changing its speed or direction arbitrarily, and traditional dynamic obstacle avoidance strategies based on object detection require massive computing resources in the prediction step, which reduces the real time characteristics of the system. To address this question, we propose a dynamic obstacle avoidance framework. It defines the environment as a random dynamic system by using dynamic grid maps, and obtains the velocity of multiple objects with low delay under GPU parallel acceleration calculations. Naturally, we introduce the relative velocity into the artificial potential field function. By expanding the data structure of costmaps, the dynamic grid map is effectively connected with the local obstacle avoidance algorithm. Our approach reduces the detour time of the robot and improves the system security.