Modeling And Optimize Control Of 3-D Gaits For Snake-like Robots In Complexity Environments

Similar to biological snakes,snake-like robots can imitate several gaits of biological snakes due to their multi degree of freedoms,slender bodies and flexible gaits.They have huge potentialities for disaster searching and rescuing.At present,the research of snake robots mainly focuses on designing mechanisms and controlling basic gaits in simple environments.However,the lack of model accuracy and adaptiveness in complex environments limit their applications in practical tasks.In order to improve their adaptiveness in complex environments and practical application in searching and rescuing tasks,this paper focuses on designing novel mechanisms,modeling of three-dimensional(3-D)gaits,optimal and adaptive controlling of gaits in complex environments.The thesis content is highlighted as below.1.Designing of snake robots'mechanisms.To improve the adaptability of snake robots in disaster terrains,a multi DOFs module composed of an orthogonal joint and a parallel joint mechanism is designed based on bionics theories and mechanisms of biological snakes.A variable stiffness actuator of snake robot is designed based on the magnetorheological damping.A snake robot composed of orthogonal joints and passive wheels is designed.Hardware such as actuators and controllers of each mechanism are selected.2.Modelling of 3-D gaits of snake robots.To improve the accuracy of mathematical models of snake robots,movements of modules in the vertica1 direction are considered.A 3-D kinematic model of a snake robot is derived by using the D-H method.A friction model is researched based on the continuous beam theory and the viscous friction model.Force and moment balance equations of the robot are derived by using the Newton-Euler method,and a 3-D dynamic model is derived.To simplified the model,actuated and underactuated DOFs are separated by using a partial feedback linearization method.3.Controlling and optimizing of snake robot gaits.Based on the movement mechanism of biological snakes,a unified gait description function of these gaits is built by using the algebraic geometry and differential geometry theories.A neural oscillator topology network is built to generate signals of all rhythmic gait.To avoid suddenly transition of gait parameters,a smooth gait transition method and a smooth start-up method are proposed in this paper.It is helpful to realize adaptive smooth control of snake robots in complex environments.To improve the robots'performances,relationships between parameters and efficiency of classical gaits are established and optimized.To improve the adaptability of snake robots,a series of novel hybrid 3-D gaits are designed by using the partial control strategy.Relationships between these gait parameters and the efficiency are analyzed and optimized to improve performances of snake robots.4.Motion planning of snake robots based on the deep reinforcement learning and the path integral reinforcement learning algorithms.A binary geometric feature map of the complex environment with multi obstacles is constructed.To evaluate the network parameters,an optimization loss function aims at shorting the path is designed by using a deep Q-learning network.An adaptive reward mechanism of multi objectives is studied and weights of the neural network are updated based on the experience playback technology.To generate passable paths in complex environments,this paper constructs an online learning framework of optimal path planning.To smooth path points of snake robots in complex environments,this paper proposes a path smoothing method that fuses a Floyd algorithm and a moving average algorithm.For the planned path points,local path segments are trained parallel to optimal their smoothness in global based on the path integral reinforcement learning algorithm.The gait control parameter space of the global path is explored serially based on the prior knowledges,an end-to-end mapping between the planned path points and gait parameters is established.Path planning and target oriented moving of snake robots are realized by using above methods.5.Adaptive path following of snake robots based on the line of sight(LOS)law.This paper analyzes the influence of key parameters on the tracking accuracy and designs an adaptive path-following algorithm for planar gaits based on the LOS law.Besides the planar gaits,a novel direction control strategy and an adaptive LOS law of 3-D gaits are designed.To improve the following accuracy,this paper designs a close-loop control frame composed of gait generating,environment measuring,direction controlling and adaptive path following.Simulations and experiments show effectiveness and adaptability of the control algorithms investigated in this paper.The control methods proposed in this paper can meet the needs of autonomous motion and gait control for searching and rescuing tasks in complex environments.