Multi-Sensor Obstacle Recognition and Fusion Method Based on Costmap

Abstract

With advancements in robotics, mobile robots are increasingly deployed in diverse applications. However, complex indoor environments pose challenges for perception and obstacle recognition. This paper proposes a costmap-based multi-sensor fusion method that integrates visual point clouds with two 2D LiDARs for obstacle detection. A downward-facing LiDAR enables ground obstacle detection, with systematic functional design and parameter optimization. Experimental results demonstrate effective 3D obstacle perception and improved obstacle avoidance, achieving an average response time of 1.09 s for static obstacles and 1.30 s for dynamic obstacles.