A vision-based real-time obstacle avoidance’s rules utilising grid-edge-depth map

Abstract

This paper presents a new rules-based of a real-time decision system for an autonomous wheeled robot with the holonomic-drive system. The robot uses decisions to avoid collisions with obstacles. The decision rules based on Grid-Edge-Depth Map. The Grid-Edge-Depth map represents the obstacle’s position and distance in the environment. The generation process of the Grid-Edge-Depth map presented in previous research. The decisions of the first scenario with no destination point are forward, stop, 90^o right turn, and 90^o left turn. The decisions of the second, and third scenarios with a destination point are forward, stop, 90^o right turn, 90^o left turn, 45^o forward to the right, 45^o forward to the left, slide to the right, and slide to the left. The proposal tested in a 5x3 metre living environment. Finally, the experiment resulted in 93.3% of navigation’s success for all the scenarios.