Research On The Multi-sensor Information Fusion Based Pose Estimation Of Mobile Robot And The Speed Control

In recent years,with the continuous progresses and developments of science and technology,related technologies of mobile robots have gradually played important roles in the actual life and production process of human beings.The related industries have imperative requirements on the capabilities of mobile robots.As the technical basis for mobile robots to complete tasks in unknown environments,autonomous pose estimation and motion control have attracted widespread attention of scholars at home and abroad for recent years.Aiming at the actual functional requirements of mobile robots in complex environments,this paper used the real-time sensing information of monocular vision sensor,IMU and wheel odometer to conduct research on the multi-sensor information fusion based pose estimation of mobile robot and the speed control.First of all,the front-end processing research on mobile robot's multi-sensor information fusion was carried out on the basis of the multi-source sensor information from monocular vision,IMU and wheel odometer.In the research on sensor information processing of monocular vision,the LK optical flow method was used to achieve the matching of Harris corner features between different visual image frames,and the method of reverse optical flow tracking was used to protect the accuracy of visual feature matching.Then,visual keyframes could be selected on the basis of visual features optical flow,and the research realized the pose estimation between visual keyframes according to the epipolar geometric constraints.In the research on sensor information processing of IMU and wheel odometer,the motion states of mobile robot were estimated by correspondent sensor information.On these foundations,the initial calibration of the motion observation scale of each sensor was carried out,providing redundant motion observation information of multi-source sensors for the process of the multisensor information fusion based pose estimation of the mobile robot.Secondly,the multi-sensor information fusion based pose estimation of mobile robot was studied according to the nonlinear optimization theory.a sliding window model of mobile robot's motion observation containing the relevant information of monocular vision,IMU and wheel odometer was constructed,using the nonlinear optimization fusion method of multisensor information.The model could realize the least square optimization of the mobile robot's motion observation residuals under the redundant motion observation information of the multiple sensors.Therefore,the position,posture and other motion states information of mobile robot could be accutately estimated by the multi-sensor information fusion method.In addition,a four-wheel differential mobile robot system based on the upper and lower computer control structure was designed.To improve the performance of the traditional mobile robot's speed control method,a mobile robot's speed control method based on optimized estimation of motion speed was proposed.In the method,according to the wheel odometer speed measurement deviation information obtained in the process of multi-sensor information fusion,the optimization estimation of mobile robot's speed estimation was realized.Then,the method could realize the mobile robot's speed control through the PID control strategy,which improved the robustness of the mobile robot's speed estimation in complex environment.Finally,the experimental platform was built according to the design plan of the mobile robot,and the pose estimation and the speed control experiments of mobile robot were carried out respectively.Relevant experimental results showed that the pose estimation method in this paper exhibited better mobile robot's pose estimation accuracy and robustness in different experimental environment.Compared with the traditional speed control method,the speed control method in this paper had better accuracy and environmental adaptability during the linear motion of the mobile robot.