A Low-Cost Environment-Interactive Patrol Inspection System With Navigation Based on Sensor-Fusion and Robotic Arm Contact Pose Feedback

Abstract

In this paper, we introduce a cost-effective mobile-robot-based patrol inspection system that navigates using sensor fusion and arm contact feedback, bypassing the need for 3D LiDAR, physical odometry, and external positioning systems. Our system utilizes a height-adjustable, four-wheel platform equipped with a 6-DOF robotic arm, achieving nine degrees of freedom with the platform’s height adjustability and planar movement. Within the Robot Operating System (ROS) framework, the system employs 2D LiDAR and a depth camera for SLAM-based mapping and pose estimation. The primary challenge during the implementation of this system is to obtain reliable pose updates of the mobile platform without physical odometry and a direct positioning source while maintaining affordability. To address this challenge, a lightweight deep neural network (DNN) object detection model is trained to identify the specific interactive items at checkpoints. By integrating a contact sensor and knowing the position of the button on the map, the acquisition of the pose of the end effector is achieved upon contact. This allows a precise update of the position of the mobile platform on the map through transforms. Experimental results indicate that our system can efficiently patrol designated routes, interact with the environment at checkpoints, and recalibrate pose using robotic arm feedback. In real-world evaluations, the system achieves a 24.35% improvement in positional accuracy and a 26.70% improvement in orientation accuracy, demonstrating its effectiveness and robustness.