Walking robots can traverse terrains inaccessible to wheeled vehicles. Moreover, their control is a complex nonlinear problem. It is therefore not surprising that the control of walking robots is a popular research area. This thesis provides the baseline of the new research at BME IIT aimed to develop a four legged robot.
The primary aim of this thesis is to develop the control algorithms for the robot. After reviewing the theoretical framework I have laid out a control strategy based on differential geometry. The verification of the algorithm needed a simulation environment, in which the optimization of the hardware configuration of the quadruped robot is also possible. One important trait of the simulation environment is that it has the same interface as the HW communication module, ensuring that simulation results can be verified on the created hardware as well.
The two main control components are the shifting of the center of gravity and the footstep planning. The latter is based on the relation between the robot's center of gravity (CoG) and its supporting polygon, in accordance with the static walking paradigm. Step target assignment is executed by the algorithm in a way that not only ensures the stability of the robot, but also takes workspace boundaries into consideration. CoG-shift is executed using velocity kinematics, while taking the size of the workspace into account as well.
Workspace analysis is concluded analytically, utilizing the static and dynamic parameters of the workspace within the control algorithm. This novel method ensures a more complete utilization of the workspace than in the case of other implementations, maximizing the agility of the robot.
Based on the simulation results the optimization of the robot's mechanical architecture is also concluded. Results show that the robot is able to reach an arbitrary -- accessible -- spatial position and orientation, even in more consecutive points.