Research On Robot Gait Control Model Based On SSD

With the continuous progress of artificial intelligence technology,path planning and motion control with target recognition algorithm is the current research hotspot of mobile robot technology,and also the key to achieve automatic cruise and complete complex intelligent tasks.At present,the target recognition algorithm of mobile robot is difficulty in meeting demands of practical application in terms of accuracy and speed.Meanwhile,the traditional control mode is relatively mechanical,which will lead to low utilization efficiency of resources in actual use.Therefore,the existing target recognition algorithm and gait control mode need to be further optimized.Based on the footstep robot as the research object,this paper adopts the improved SSD and the principal axis rotation method to establish the obstacle feature extraction and target recognition model and the ADRC control algorithm,so as to realize the rapid and accurate classification of obstacles.Research on the gait control model of robot based on SSD,the main research contents and innovations include:(1)An obstacle recognition algorithm based on improved SSD(Single Shot Multi Box Detector)is proposed.The soft-NMS(softening non-maximum suppression)is used to optimize and improve the network structure,and the anchor box are acquired to provide better initial values for bounding box prediction.Pooling layer adopts the space pyramid pooling(SPP)method to replace pooling layer,keep the image size unchanged and reduce the over-fitting.The gradient is made more stable and predictable by Batch Norm so as to speed up the constriction and training velocity of network learning.The recognition precision of the improved SSD algorithm is 89.5%,and the recognition velocity is 38 frames per second.Compared with the traditional SSD algorithm,the precision is improved by 3.8% and the velocity is improved by 2 frames.(2)A method of obstacle size measurement based on principal axis rotation is proposed.When robot encounters the type of obstacle with large height change,the improved SSD algorithm is used to classify the obstacles,and then the height of the obstacles is measured based on the improved principal axis rotation.It provides three types of gait control mode for robot,including ‘walking along the original road without changing the gait,improving the chassis to cross obstacles and changing the path to avoid obstacles'.(3)An improved gait control model based on target recognition obstacle classification results is proposed.According to two times classification results,the active disturbance rejection control model is used to build the robot experimental platform.Three kinds of gait and trajectory of the robot are proved when it encounters obstacles.The results display that the improved SSD algorithm can improve the performance of gait control and path planning,and the accuracy reaches 83%.