Human Motion Capture And Behavior Sample Generation And Experimental Research Based On PN Pro

With the optimization of mechanical structure,increasing in number of Do F and types of sensors,the interaction with humanoid robot could be achieved by means of voice,image,data and action.It was supposed that humanoid robots could be applied broadly on entertainment,medical,military and disaster relief and other field.Once it came true,humanoid robot would help people to accomplish the work,improve efficiency and ensure the health and safety.Walking stably is the premise for robot which could be applied on the mentioned fields.Because the shape and structure of humanoid robot were designed by the structure of human body,the law of human behavior was used to drive the humanoid robot for the purpose of making the behavior between human and humanoid robot similar.Once it implemented,the robot could not only walk stably but also own the characteristics of human walking.Therefore,researches on how to use inertial motion capture equipment capturing human motion data,form human motion samples,and apply motion law of human on humanoid robot were carried out.First of all,the inertial motion capture suit,Perception Neuron PRO,was used to build a data acquisition system for human motion,which was able to capture,extract,process and store motion data.Consequently,the data of human daily behavior such as walking,jumping,squatting,sitting down,standing up,going to bed,getting out of bed and lifting objects was captured and transformed into behavior samples.Secondly,the captured data of human walking was processed.According to the gait parameters which was calculated and the motion characteristics which was summarized by the motion trajectories from lower limb joints and the Co M,the motion laws of walking were obtained.By applied these rules on humanoid robot,the gait of the robot could become anthropomorphic.In order to complete the transformation from human behavior samples to robot behavior samples,the motion of Co M and the swinging foot,which belonged to human motion space originally,were mapped into the robot action space.The desired trajectory of robot Co M and swinging foot could be obtained by proportional change.By solving the forward kinematics of the robot,the real trajectory of the Co M and the swing foot which was represented by the angle of robot lower limb were obtained.After these,a nonlinear optimization problem under the constraints was proposed to get angle of robot lower limb.The similarity of the rotation matrix between the robot and the human lower limb was regarded as the objective function,and the range of joint motion,the attitude of the swinging foot when landing,walking stability and the collision avoidance were regarded as the constraints.The Sequential Quadratic Programming was used to search the optimal solution of the optimization problem.The walking samples of the robot were obtained.Finally,an experimental platform with Nao robot as hardware was built to verify if the walking of humanoid robot could be driven by robot walking samples.The results proved the rationality and validity of the robot behavior samples which were transformed from human behavior samples.