A Vision-Based Approach for Autonomous Motion in Cluttered Environments
Zhenping Wu,Zhijun Meng,Yulong Xu,Wenlong Zhao
DOI: https://doi.org/10.3390/app12094420
2022-01-01
Applied Sciences
Abstract:In order to complete various tasks automatically, robots need to have onboard sensors to gain the ability to move autonomously in complex environments. Here, we propose a combined strategy to achieve the real-time, safe, and smooth autonomous motion of robots in complex environments. The strategy consists of the building of an occupancy grid map of the environment in real time via the binocular system, followed by planning a smooth and safe path based on our proposed new motion-planning algorithm. The binocular system, which is small in size and lightweight, can provide reliable robot position, attitude, and obstacle information, enabling the establishment of an occupancy grid map in real time. Our proposed new algorithm can generate a high-quality path by using the gradient information of the ESDF (Euclidean Signed Distance Functions) value to adjust the waypoints. Compared with the reported motion-planning algorithm, our proposed algorithm possesses two advantages: (i) ensuring the security of the entire path, rather than that of the waypoints; and (ii) presenting a fast calculation method for the ESDF value of the path points, one which avoids the time-consuming construction of the ESDF map of the environment. Experimental and simulation results demonstrate that the proposed method can realize the safe and smooth autonomous motion of the robot in a complex environment in real time. Therefore, our proposed approach shows great potential in the application of robotic autonomous motion tasks.