Research On Cognitive Learning System Of Lateral Balance Control Of Unmanned Bicycle

In the self-balancing control of unmanned bicycle,it's the basis research that control lateral balancing.In the balance control of unmanned bicycle the traditional controller has shown well control effect.However,when it comes to a changeable and complex environment,the traditional controller needs to readjust its control parameters.At this time,the stability and robustness of the controller can not be suit for the actual needs.Inspired by intrinsic motivation in bionics,reactive cognitive learning system is introduced into unmanned bicycles,so that unmanned bicycles can interact with the environment continuously,stimulate the intrinsic motivation of unmanned bicycles through learning mechanism,make the system reach a balanced and stable state,and determine the nonlinear mapping relationship between system output actions and system state characteristics.In order to stimulate the intrinsic motivation in the process of interacting with the environment and improve the self-learning ability and environmental adaptability of unmanned bicycles through reactive cognitive learning,the main research contents are as follows:1)Analyze the dynamic model of unmanned bicycle.Choose the suitable dynamic model of unmanned bicycle for reactive cognitive learning research.A nonlinear second-order response mechanical model which can characterize the dynamics of unmanned bicycle is established by adopting Chaplygin equation,and the model is transformed to obtain a linear model.Then introduce a linear variable parameter(LPV)model of unmanned bicycle.Compare and analyze the strengths,weaknesses and applicability of these three dynamic models,and the LPV model of unmanned bicycle is selected as the dynamic model for this study.2)According to the physical prototype used in this research,analyze and study the acquisition of characteristic parameters in LPV model.The characteristic parameters of LPV model of unmanned bicycle are acquired by offline calculation and online recognition,and then two definite models are obtained.Compare and analyze the dynamic responses of the two models,then select the model which is closest to the dynamic response of the physical prototype to be used for researching reactive cognitive learning system of unmanned bicycle.3)Introduce the reactive cognitive learning system of unmanned bicycle.Based on a reactive cognitive learning system,a cognitive learning mechanism based on probability learning is constructed to realize self-learning of lateral balance control of unmanned bicycle.Counting on the dynamic model of unmanned bicycle and matlab/simulink module,the learning system model is constructed and the self-balancing simulation learning of unmanned bicycle is expanded through the learning mechanism.By analyze the results,the self-learning of lateral balance of unmanned bicycle can be realized by adopting the reactive cognitive learning system,and the "state-action" rules obtained by learning can control the lateral balance of unmanned bicycle well.4)Carry out the physical prototype experiment of reactive cognitive learning of unmanned bicycle.Introduce the physical prototype of unmanned bicycle and its measurement and control system.Based on the simulation learning foundation,carry out the experiment based on reactive cognitive learning on the physical prototype.Driven by cognitive learning,the physical prototype can interact with the environment continuously,and make independent action selection and action learning.By analyzing the results,it is found that the physical prototype of unmanned bicycle can stimulate the intrinsic learning motivation of unmanned bicycle and make it emerge from learning ability.