Virtual model-based control method for quadruped robots

Abstract

In this thesis, the focus is on the walking planning, analysis, and control of a quadruped mobile robot. The gait patterns of quadrupedal animals were thoroughly examined, and as a result of this examination, the Trot Gait was chosen due to its adaptability to challenging terrain conditions, wide speed range, and ease of balance control. This gait pattern is frequently preferred in both quadrupedal animals and robots. The Trot Gait involves the simultaneous movement of two diagonal legs of a robot (for example, the right front and left rear or left front and right rear). The decision on which two diagonal legs are in swing mode and which two are in stance mode during the control of the quadruped robot is determined by an Event-Based Finite State Mechanism. This mechanism allows the robot to dynamically adapt to environmental changes and different ground conditions. Generally, quadruped robots possess a complex and nonlinear dynamic structure, being either fully actuated or underactuated. Consequently, control methods based directly on these dynamic equations involve a high level of computational complexity. The primary objective of this study is to propose a simpler and more real-time computation-friendly method using the Virtual Model Control (VMC) approach. In this approach, the effects of the coefficients of critical virtual spring and damper components on robot control have been examined. These virtual components play a significant role in the robot's adaptation to complex terrain conditions and in exhibiting stable behavior. Simulation results have demonstrated that the appropriate selection of virtual spring and damper parameters significantly benefits the walking balance, environmental adaptation, and error reduction in following given references for the quadruped mobile robot. The dynamic modeling and simulation of the robot were carried out with the integration of MATLAB\Simulink and MuJoCo. MuJoCo's advanced physics engine has facilitated the detailed and realistic simulation of dynamic processes, especially in studies where environmental interactions are prominent. The control and simulation of the quadruped robot were achieved using feedback signals from the robot modeled in the MuJoCo environment and control signals calculated by the VMC method developed in MATLAB\Simulink. The obtained results demonstrate that the VMC method can be effectively used in the motion control of quadruped robots.