Study On The Telepresence Control Of Humanoid Robot With Full-body Movement Via Brain-robot Interaction

Brain-robot interaction(BRI),which provides a bidirectional communication pathway between human brain and a robotic device via brain-computer interface and sensor feedback,is an interdisciplinary study of important scientific value and application prospect.On the basis of integrating two service technologies of BRI and humanoid robot,the telepresence-controlled humanoid robot based on BRI,also named brain-controlled humanoid robot,fuses brain and machine at perception and decision-making level by fully exploiting the cognitive advantage of the human brain and the autonomous intelligence of the robot system.It is not only prospective in assisting the sick and elderly,but also of important value in military fields.In addition,this study is helpful to deepen the understanding of brain mechanism in advanced cognitive process and promote the development of multiple research fields such as neuroscience,signal processing,robotics and artificial intelligence.Considering the key issues in the brain-controlled humanoid robot research,this study makes the following contributions which are also the major innovations of this thesis.1)Based on unified platform and standards,the performance of steady-state visual evoked potential(SSVEP)and P300 based models for the on-line telepresence control of humanoid robot are systematically evaluated and compared.This study first established Cerebot – a mind-controlled humanoid robot platform,and implemented both SSVEP and P300 models by configuring their key parameters according to the unified standards and experimental verification,and using the representative single-channel processing algorithms which are commonly used in current research.This study then selected subjects with diverse experience in BRI experiments and compared the on-line performance of SSVEP and P300 models for telepresence control of humanoid robot through both on-line open-loop and closed-loop evaluations.The experimental results revealed that the SSVEP model yielded faster response to the subject's mental activity with less reliance on channel selection,whereas the P300 model was found to be suitable for more classifiable targets and required less training.2)To establish the BRI system of brain-machine fusion,an SSVEP-based hierarchical architecture is proposed based on the similarities of processing information between human brain and machine.The architecture consists of three layers: the Brain-Computer Interface layer proposes an asynchronous decoding algorithm of SSVEP signals based on canonical correlation analysis(CCA)and fast Fourier transform(FFT),and develops an adaptive content-coding interface to present the visual stimuli;the Behavior Mapping layer uses a menu-based behavior mapping mechanism to coordinate multiple robot behaviors;the Robot Controller layer develops an autonomous response model to cope with emergencies.The experimental results showed that the SSVEP-based hierarchical architecture enabled the subject to control 20 behaviors of humanoid robot with full-body movement using only 5 SSVEP stimuli,and balanced navigation time and collisions better in the complex environment through flexibly switching between Walking with Fine Tuning(WFT)behaviors and Walking with Gross Movement(WGM)behaviors.3)A comprehensive evaluation criterion for brain-controlled humanoid robot system is proposed by considering its task requirements and characteristics in assisting people with movement disorders.The evaluation procedure includes three steps: the off-line calibration stage selects five frequencies with the most responsive SSVEPs for each subject;the on-line open-loop evaluation stage measures the accuracy rate,response time and information transfer rate for each stimulation frequency,and allocates these frequencies to the five visual stimuli,which encode different robot behavior,by considering the success rate and the response time of each subject;the on-line closed-loop stage designs an evaluation task by combining three typical applications of the brain-controlled humanoid robot,conducts three types of comparative experiments to perform the task using the WFT behaviors,the WGM behaviors and both WFT and WGM behaviors respectively,and calculates navigation time,number of control commands,and number of sideswipe collisions and obstacle collisions for evaluating the brain-controlled humanoid robot system in assisting people with movement disorders.