Research On Visual Perception And Gait Planning Of Lower Limb Exoskeleton

As a wearable device to assist patients with walking dysfunction in lower limb rehabilitation,lower limb exoskeleton robots have made a significant contribution to medical rehabilitation and other fields in recent years.The coordinated control of lower limb exoskeletons relies on a variety of key technologies,among which gait planning technology is used to achieve exoskeleton gait trajectory generation,which is one of the technical difficulties in the field of exoskeleton control.Exoskeletons that use fixed gait trajectories for locomotion are inflexible and have significant limitations.In addition,most of the current exoskeleton robots lack the ability to actively sense the walking scene.This thesis addresses the environmental perception and gait planning techniques for exoskeletons,with the following main research elements:(1)Aiming at the problem that the existing lower limb exoskeleton robot lacks the ability to recognize walking scenes and the low accuracy of parameter estimation for stairs with special materials,this study gives the exoskeleton the ability of visual perception by adding a depth camera to the exoskeleton and proposes a robust stair parameter estimation method based on plane detection for the point cloud generated by the depth camera.In this study,the scene recognition of RGB images acquired by the depth camera is performed by constructing a Rep VGG model.In the scene detected as a staircase,the staircase point cloud is estimated by the staircase parameter estimation based on line contour recognition and the robust staircase parameter estimation algorithm based on plane detection,respectively.The experimental results show that the plane detection-based robust staircase parameter estimation algorithm,which can use both RGB color and normal vector information of the point cloud,achieves better parameter estimation accuracy compared with other parameter estimation methods with recognition errors of 2.77% and 5.96% for the width and height of the four staircase scenes,respectively.(2)Aiming at the problem that the existing lower limb exoskeleton robot stair climbing scheme cannot switch according to the scene and cannot be adaptively adjusted according to different size steps,this paper proposes a gait planning method for exoskeleton stair climbing that can adapt to scene switching and different size steps.In this study,a trajectory planning model based on over-point motion primitives is constructed,and the collected gait trajectory curves are input to the model as source trajectories for training.According to the recognition results of the scene recognition algorithm,the exoskeleton going up the stairs is divided into three stages for gait planning respectively.In each stage,key points are set according to the step parameters and gait parameters,and the gait trajectory is planned and modulated by the over-point motion primitive model.Finally,in the process of gait planning model evaluation experiments,this paper verifies the effectiveness of the proposed gait planning method for walking up stairs.The experimental results show that the proposed gait planning method can produce accurate trajectories,and its error with the actual running trajectory of the human body after wearing is about 1cm,which is effective.(3)To address the problem that the existing lower limb exoskeleton upper computer system lacks visual perception and gait information management functions,a system is designed and implemented in this study.In terms of motion control,the system sends control commands to the lower limb exoskeleton through a combination of wired serial port and wireless WIFI,and the whole system mainly contains two functional modules: visual perception and gait information management.In the visual perception module,the system recognizes the current walking scene of the exoskeleton in real time through the depth camera,and identifies the parameters of the staircase scene.In the gait information management module,the user’s personal information of the exoskeleton user can be managed,and the gait trajectory that best fits the personal characteristics of the current user can be generated.