Safety Evaluation Method Of Robot Transfer Based On Multi-source Signal Fusion

With the rapid development of artificial intelligence industry in China,robots are widely used in daily life.For the disabled and the elderly who are inconvenient to move,it is convenient for many families to operate nursing robots to realize simple daily life behaviors,which greatly reduces the burden of medical staff.In the process of getting up and taking a seat transfer assisted by a wheelchair robot,In the process of rising-sitting down transfer assisted by robot,it is necessary to evaluate the safety of human-robot interaction process,so as to guide the safe transfer behavior.Therefore,this thesis develops a rising-sitting down transfer planning method with strong generalization ability for different mobility capabilities users.After analyzing the difference of pressure and brain oxygen signal under different migration behaviors,a safety evaluation model of robot transfer based on multi-source signal fusion can be established,which can improve the safety of human-robot interaction,and provide important guidance value for the daily work and service quality of medical staff,the elderly and people with different behavioral abilities.Based on the daily behavior of getting up and taking a seat transfer of a wheelchair robot operated by a person,the movement characteristics of the human body in the process of sitting down are analyzed under the conditions of human-robot cooperation and normal sitting,a threedimensional model of human body sitting down is proposed,which is more suitable for the actual situation.The mapping table of human body position and sitting point is established by using the model of human body sitting down and the physical dimension of the robot.Through the laser sensor to obtain the human leg information and matching with the mapping table,the moving target of the robot can be obtained.On the basis of the method of rising-sitting down transfer based on the calculation of the moving points,combined with the reinforcement learning method,after the robot moves to the designated position and the human body is seated,the multi-source signals are fused by artificial neural network to predict the safety evaluation value of the transfer state,and the reward required by reinforcement learning is given to build the transfer safety evaluation model.According to the results of safety evaluation,the self-learning strategy of sitting point is formulated,and the sitting point compensation is carried out for the established three-dimensional human sitting model,and the rising-sitting down transfer plan based on the reinforcement learning is proposed,so as to realize the safe rising-sitting down transfer.To verify the validity and accuracy of the transfer method proposed in this thesis,an experimental platform is built based on the omni-directional mobile intelligent wheelchair robot independently developed by our research group,and the simulation analysis and the transfer experiment are carried out respectively.The experimental results show that the average pressure of back and buttock is reduced by 5.05% and 3.95% respectively compared with autonomous behavior through the method of rising-sitting down transfer based on the calculation of the transfer points proposed in this thesis,the instantaneous impact force is reduced by 4.61% and22.56% respectively,and the contact area is increased by 11.03% and 2.67% respectively.Through the method of rising-sitting down transfer based on reinforcement learning,the intelligent wheelchair robot can obtain the corresponding safety evaluation according to the characteristics of different users,and adjust the movement strategy according to the evaluation results to optimize the transfer points.The evaluation results predicted by the safety evaluation system are highly close to the subjective evaluation,which verifies the accuracy of the safety evaluation method proposed in this paper.After the reinforcement learning thought is introduced to optimize the sitting point,the subjective safety evaluation value and its predicted value are significantly improved compared with those before the sitting point optimization,and the actual value of the safety evaluation of transfer can be improved by 33.33%,thus verifying the improvement of safety by the transfer method proposed in this thesis.