Dynamic obstacle avoidance for efficient exploration of ground robots

Abstract

Collision avoidance for moving obstacles is essential for autonomous robot to explore effectively in unknown dynamic environment. In this paper, we propose a collision avoidance module that can avoid moving obstacle during the exploration. Moreover, we minimize the cost of efficiency resulting from avoiding maneuvers. While existing solutions assume prior knowledge of the obstacle's movements, the proposed method derives such information based on observation. The proposed method first detects moving obstacles and extracts velocity information. Then it estimates the expected collision point and modifies the map used by navigation. Lastly the revenue scores for each goals points are updated based on the newly calculated paths. Thus the proposed method is capable of avoiding moving obstacles and minimize the time and distance traveled for exploration. A simulator based experiment is conducted to compare the proposed algorithms with control methods. The results shows that our method outperforms the other existing methods in terms of time, distance and number of collision.