Research On Cane-Type Walking-Aid Robot For People With Mild To Moderate Lower Extremity Muscle Strength Deficits

Decreased muscle strength in the lower extremities due to aging,neurodegenerative diseases,and sports injuries severely limit the abilities of patients to perform daily living activities.After treatment and rehabilitation,at least 70% of patients regain partial motor function.In order to encourage people with moderate or mild lower limb muscle weakness to actively participate in rehabilitation training,the cane-type walking robot focusing on“assist as needed” can be used as an effective alternative and supplement to physicians and caregivers by providing the walking assistance in assistive training,the human-following in independent training for people with moderate to mild lower extremity muscle strength deficits,respectively,as well as the gait analysis system for physicians to assess the effectiveness of rehabilitation.It has become a hot research topic in the field of walking-aid robots.Although walking-aid robots have been extensively studied,there are still many challenges in practical applications.In the walking assistance,it is difficult for the robot to accurately extract and quantitatively describe the human motion intention,and it is difficult to accurately recognize diverse walking states.In the human-following and supervision,it is difficult for the robot to obtain human sagittal information in real time,and it is difficult to realize the non-contact human-following within the fixed human-robot relative posture.In the gait analysis,it is still difficult to accurately extract gait parameters based on mainstream research of onboard sensors.In order to solve these challenges,in this dissertation,the mechanical design and software system of multi-function integration of a cane-type walkingaid robot are studied for meeting the diverse needs of people with moderate and mild lower extremity weakness,which realize the human-robot coordination-based walking intention fusion estimation algorithm,the data-driven based human-robot coordination walking state monitoring,the human-following control with fixed relative posture as well as the gait spatiotemporal parameter extraction algorithm based on the onboard laser range finders(LRFs)are proposed.The contributions are summarized as follows.To solve the problem that the current walking-aid robots have single function and are difficult to meet the diverse needs of people with moderate and mild lower limb muscle weakness,combined with the actual use scenario of the robot in this dissertation,for the high mobility needs of diverse environments,the lightweight mechanisms such as a miniaturized omnidirectional mobile chassis and a resin 3D printed handle are designed by using the weight reduction design method.Based on the functional requirements of walking assistance,human-following and gait analysis,a multi-level system of “perception,intention estimation and robot control” is constructed.The kinematics model of the cane-type walking-aid robot is established,which provides a theoretical basis for the robot control.To solve the problem that it is hard to accurately identify the user’s walking intention and walking state in walking assistance based on the robot,the human-robot coordinationbased walking intention fusion estimation algorithm is proposed in this dissertation.Guided by real-time estimation of walking intention and combined with the robot admittance control technology,a compliance and smooth walking assistance control is achieved.The experimental results show that the proposed method significantly improves the smoothness index of human-robot interaction compared with the walking assistance based on the traditional force-based robot control,with an average improvement of 52%.Based on the hypothesis of human-robot coordination and the fault diagnosis principles of principal component analysis,the data-driven human-robot coordination walking state monitoring algorithm is proposed in this dissertation,which can effectively identify the normal walking state,abnormal walking state with limb constraints as well as emergency state within 55～110 ms.Compared with the conventional fall detection method based on the center point of pressure,it can detect the emergency state more quickly and effectively.For the problem that it is hard to extract the human positive orientation angle accurately and walking intention in the human-following for the robot,a unified model is proposed in this dissertation to describe the evolution of human intention quantitatively.Moreover,an online walking intention estimation algorithm is proposed.The average measurement errors of intention position estimation in two-dimensional directions of the horizontal plane are 9.4cm and 9.2 cm,respectively.The average measurement error of intention facing orientation estimation is 13.1?.In order to ensure the satisfied human-following performance,a finitetime human-following control algorithm with fixed relative posture is proposed,which can guarantee the average following errors in two-dimensional directions of the horizontal plane as well as the intention facing orientation are 7.9 cm,6.9 cm and 12.1?,respectively.The performance is better than the human-following control methods based on the PID,virtual spring and model predictive control algorithms.To solve the problem that it is difficult to accurately extract 3D gait spatiotemporal parameters based on only two onboard LRFs,a gait analysis algorithm based on high-dimensional Takagi-Sugeuo-Kang fuzzy system is proposed in this dissertation.The walking data acquired by LRFs are labeled using the plantar pressure sensor as the golden rule for the gait event detection.Furthermore,the gait data set is constructed for training the highdimensional Takagi-Sugeuo-Kang fuzzy system.The gait cycle is divided based on the results of walking state classification to achieve accurate extraction of gait parameters.In addition,based on physician experience and the Tinetti scale,the walking ability evaluation index and robot online evaluation algorithm are proposed.The clinical experiments show that the proposed high-dimensional Takagi-Sugeuo-Kang fuzzy system based gait analysis algorithm achieves an accuracy rate of 96.57% for walking state recognition and exemplary performance in gait parameter extraction,with an average measurement error of 0.02 m in the step length and 0.13 s in the gait cycle.Besides,the walking ability assessment results are consistent with the scores given by the rehabilitation physician.