Research On Motion Intention Recognition Technology For Lower Limb Exoskeleton Robot

The lower limb exoskeleton robot is a highly mechatronic wearable auxiliary device with bionic structure,which has a wide range of application prospects in medical,industrial and military fields.Exoskeleton robots involve technologies in sensor information processing,motion intention recognition and intelligent control as a complex human-machine interaction system.Human motion intention recognition is one of the crucial steps of exoskeleton-assisted human motion.At present,many exoskeleton systems are controlled based on predefined gait or patterns.This approach ignores the initiative and enthusiasm of human movement and cannot accurately identify human motion intention.There are some problems such as uncoordinated movements and exoskeleton dragging wearer to walk,increasing the walking burden of wearer.This paper carries out research on motion intention recognition technology for lower limb exoskeleton robot aiming at the problem that exoskeleton cannot accurately recognize human motion intentions and is difficult to realize coordinated humanmachine motion.The research focuses on critical technologies such as gait phase recognition and motion pattern identification of human body.The lower limb exoskeleton experimental platform is built to verify the effectiveness of proposed algorithm.The research in this paper lays a technical foundation for active motion control based on motion intention recognition in lower limb exoskeleton robots.The main research of this paper is as follows:(1)The human motion gait is analyzed,including the analysis of lower limb motion mechanism,and the division of human gait phases and motion patterns.The types and characteristics of sensors commonly used in lower limb exoskeleton robots are investigated by referring to relevant literature.Inertia measurement unit is selected to construct the motion signal perception system.The design scheme of control system and actuator is described.The STM32 controller and motor driver are selected together to form the control system.The DC brushless motor is selected as actuator of exoskeleton.Finally,it is introduced the structure and function of the developed lower limb exoskeleton experimental platform.(2)The gait phase recognition method based on integrated network model is proposed in this paper.The model can accurately identify four phases in the gait cycle,including heel strike,foot flat,heel off and swing phase.Normalization and feature extraction of collected sensor signals are performed to enhance the accuracy of recognition during the gait identification process.The processed data are input into the integrated network model.The introduction of sparse autoencoder into the model can extract key features from gait data.Bidirectional long short-term memory is used to learn dynamic change patterns within features.Deep neural network is adopted to identify gait phases and output classification results.The integrated network model is compared with the basic models for experiments.The results demonstrate that the proposed model is much more accurate than other models for gait phase recognition.(3)The gate recurrent unit(GRU)incorporating multi-head attention mechanism is constructed to identify different motion patterns.GRU can capture periodic changes from time-series data in the GRU-MHA model.Multi-head attention mechanism enhances the influence of significant feature information output from GRU in the overall input sequence.The collected data are divided into steady pattern dataset and transformed pattern dataset according to different experimental processes.The proposed model is tested on different datasets respectively.The results show that the GRU-MHA model outperforms other comparative models in motion pattern identification,and can accurately recognize five steady motion patterns and eight transformed patterns for a total of thirteen motion modes.This paper focuses on the motion intention recognition technology for lower limb exoskeleton robot.The lower limb walking-assisted exoskeleton robot is developed for disabled people.The gait phase recognition algorithm and motion pattern identification algorithm are designed,which can effectively recognize the motion intention of wearer and lay the foundation for realizing the fine control of exoskeleton system.