Study On The Planning And Stability Of Omnidirectional Locomotion For Quadruped Robot

When developing the legged robot, four-legged mammals gain its popularity in researchers and become the best sample of imitation for their powerful locomotion ability and stable maneuverability. Thanks to developments in technologies, such as hydraulic servo, sensor detection, bionic control, hydraulic actuated quadruped robots start to come on stage frequently. These quadruped robots, with the characteristic of good adaptability, wide movement range and heavy load capacity, are perfect mobile platforms for implementing missions of mountain transportation, disaster relief, and military reconnaissance. Thus, they are of high expectations and hope. However, being able to perform flexible, stable, reliable and perdurable locomotion in a wild world is the precondition of accomplishing aforementioned missions and also the hot research topic in upto-date research in quadruped robots. To improve locomotion ability of quadruped robot in the wild world, this thesis will take a hydraulic quadruped robot newly developed by our team as the research object, and focus on several issues related to its omnidirectional locomotion, which is to conquer core technical problems and master the corresponding key technologies. The specific work and main contributions are as follows:(1) The hydraulic quadruped robot studied in this thesis is introduced from the aspects of mechanical structure, hydraulic system and sensing control system. Firstly, the bionics design process of the mechanical structure of the robot is described in this chapter. Secondly the schematic diagram of the hydraulic system is given to illustrate the working principle and parameters of the integrated oil tank. Thirdly the controller and sensors equipped on the robot are introduced. Finally, the final physical prototype of the robot is showed for further discussion in the following contents.(2) The omnidirectional locomotion of the trot gait for the quadruped robot is studied in this chapter. To begin with, the omnidirectional locomotion is regarded as a combination of steering and oblique locomotion, and then the locomotion parametric matrix method is proposed to plan the steering and oblique locomotion of the trot gait. Inspired by quadruped animals, the centroid offset adjustment stabilization algorithm is proposed to improve its locomotion stabilization. The policy gradient via the signed derivative(PGSD) method is used to determine the optimal distance of centroid offset. Then the Fourier series fitting function is used to fit the optimal distance of different steering angle or oblique angle, which can reduce the duration of learning process. The centroid offset adjustment stabilization algorithm can also be used to realize the steering and oblique locomotion of the quadruped robot on the flat slope. Finally, the kinematic simulation is conducted to verify the effectiveness of the algorithm.(3) An algorithm for omnidirectional locomotion planning on irregular terrains is proposed. Such algorithm which is based on floating base kinematics can plan the motion displacement and angle of the robot body automatically according to the landing positions of the feet, which can guarantee the realization of planned body motion displacement during the walking phase and then of body motion angle during the adjustment phase. Overall, the quadruped robot can adjust the state of locomotion continuously according to different terrains. Finally, the quadruped robot is tested by climbing multi-level stairs in simulation to verify the effectiveness of the proposed algorithm.(4)The terrain reconstruction which based on stereo vision is studied in the thesis, and an algorithm using mobile data selection windows is proposed to accomplish lacuna plane reconstruction. To facilitate the implementation of this algorithm, the three-dimensional point cloud obtained from stereo cameras is processed by gridding. After gridding, mobile data selection windows with different sizes traverse through the gridding area. Then, according to the imaging characteristics of the plane in the stereo camera, decide whether the existing data points within the mobile data selection window are on the same plane. If the existing data points within the mobile data selection window are considered to locate in the same plane, then new data fitting the resulting plane equation are introduced into the empty grid points until there are no new reconstruction points left. The process of stair surface reconstruction is regarded as an example to illustrate the steps of the proposed algorithm and verify the effectiveness of the proposed algorithm.(5) The fault-tolerant gait planning method for quadruped robot with locked joints is discussed in the thesis. The redundancy equation of virtual degree of freedom related to the mal-functional joint is established based on the floating base kinematics. The known and unknown parameters of the redundancy equation are determined according to omnidirectional locomotion planning algorithm which is applied to irregular terrains. The realization of known parameters’ planning is guaranteed by tuning the unknown parameters. In other words, some virtual freedoms are sacrificed to ensure the realization of other more important freedoms. To solve the redundancy equation, the related optimization function is introduced into the extended Jacobian matrix for realizing the optimal control of the fault-tolerant gait. Then the effectiveness of the proposed algorithm is verified by simulation.Finally, the physical prototype experiments during the development process of the hydraulic quadruped robot are illustrated in the thesis. Two legs experiment, squat experiment and suspension gait experiment are all included. And parts of algorithms proposed in the thesis are also verified on the hydraulic quadruped robot.