Self-walking Planning Of A Hexapod Walking Robot Based On Vision

Legged robots have excellent terrain adaptability,they are able to walk in harsh environment to accomplish detecting,operating and rescuing tasks instead of human beings.In such application environments,legged robots must have the ability to perceive the environment,classify the terrain and plan motion autonomously.This paper proposes the self-walking planning of a Hexapod walking robot based on vision.The Hexapod walking robot is designed using a special Parallel-Parallel(P-P)concept,it has high stability,high precision and great carrying capacity.This paper mainly focuses on four points: the establishment of the robot pose model,the relative pose identification between the 3D vision sensor and the legged robot,the walking planning in complex environment and the walking planning on complex terrain.Multiple experiments have been carried out to verify the planning theory in our work.The major contents are listed as follows:(1)The mathematical relationship between the robot pose and its actuators is modeled based on the position model of single leg.Two based important problems have been solved:body pose is known to solve needed actuators' position and actuators' positions are known to solve the body pose.Then the Adams simulation model is used to verify the robot pose model.(2)A novel relative pose identification method between the 3D vision sensor and the legged robot is proposed.This method only needs the robot move in space without external measuring equipment and human intervention.In this method,the relatively flat ground is innovatively selected as a target.Then the point cloud is processed to obtain the ground pose with respect to the 3D vision sensor,and the robot pose with respect to the ground is calculated using the robot pose model.Based on the relationship between the robot,the ground and the vision sensor,a nonlinear optimization model consisting of identification parameters is established.The LM algorithm is used to solve the optimization problem.In the end,several real experiments are carried out to obtain the relative pose parameters and test the method's robustness.(3)The safe trajectory planning of the Hexapod walking robot in complex environment is studied.Firstly,the obstacle detection and location algorithm is proposed.The ”quaternion bucket” is introduced to describe the obstacle and the robot,which facilitates the distance computation between the robot and the obstacle.Then the virtual obstacle is proposed to avoid the robot walking on undetected terrain.And the safe conversion pose of the robot in the complex environment is defined.In the end,the third order Bezier curve is used to parameterize the walking trajectory,and the optimization method is used to solve the shortest safe path.(4)The agile trajectory planning method of the Hexapod walking robot in complex environment is studied.The searching model of the optimum walking direction in complex environment is established based on the legged robot's all-direction mobility.Considering the planning method's practicability,firstly several discrete directions are defined previously to act as candidates for the robot to select.Then assuming that the robot walks towards a specific direction for a fixed period of time,the object function is computed.After that,the optimum value of the object function can be obtained,and the corresponding optimum walking direction is selected.(5)The escaping trajectory planning of the Hexapod walking robot in complex environment is studied.Firstly,the obstacle memory method is proposed based on the ”quaternion bucket” model.Then a set of safe positions,which are far from obstacles,are calculated using the memorized obstacles' information.In the end,the segmented third order spline cures are used to generate a smooth escaping trajectory,which allows the robot to walk back safely and continuously without encountering with obstacles.(5)The walking planning method of the Hexapod robot on complex terrain is studied.Firstly,the GFGM(Geometric Feature Grid Map)is modeled to describe the real terrain.Based on the GFGM,several typical structured terrains are classified and the corresponding geometric parameters are identified.Then the foothold selection method and the gait parameters determination method based on the geometric feature of the terrain are discussed.In the end,the pre-adjustment method is introduced to maintain the robot's stability and safety during walking on complex terrain.The planning theory is verified by several meaningful and practical experiments.