Research on slipping-pattern recognition and positioning error compensation of wheeled robot

Abstract

The autonomous navigation and positioning system is one of the key equipment in industrial automation production. Due to slippage, the odometer (OD) data cannot reflect the real rotation of the wheeled robot, which may cause the unmanned robot to deviate from the preset route. To solve this problem, the integration of the strap-down inertial navigation system (SINS) and the OD is used, and a novel slipping recognition algorithm based on neural network learning and an error compensation method is proposed to improve the navigation accuracy when skidding. First, the output data of the SINS and OD are studied and features are extracted for neural network learning. After training and modeling, the state of the robot slipping can be identified quickly and effectively. Second, different models are designed to compensate the errors of the different slip patterns. Experimental results show that the comprehensive recognition rate of five modes within 1 s is better than 98.9%. After compensation, the positioning errors of the left and right slip modes have been reduced to 15 mm and the positioning errors of the both side and locked slip modes have been reduced to 40 mm. Compared with before compensation, the errors are reduced to less than 6%.