Support polygon in the hybrid legged-wheeled CENTAURO robot: modelling and control

Search for the robot capable to perform well in the real-world has sparked an interest in the hybrid locomotion systems. The hybrid legged-wheeled robots combine the advantages of the standard legged and wheeled platforms by switching between the quick and efficient wheeled motion on the flat grounds and the more versatile legged mobility on the unstructured terrains. With the locomotion flexibility offered by the hybrid mobility and appropriate control tools, these systems have high potential to excel in practical applications adapting effectively to real-world during locomanipuation operations. In contrary to their standard well-studied counterparts, kinematics of this newer type of robotic platforms has not been fully understood yet. This gap may lead to unexpected results when the standard locomotion methods are applied to hybrid legged-wheeled robots. To better understand mobility of the hybrid legged-wheeled robots, the model that describes the support polygon of a general hybrid legged-wheeled robot as a function of the wheel angular velocities without assumptions on the robot kinematics or wheel camber angle is proposed and analysed in this thesis. Based on the analysis of the developed support polygon model, a robust omnidirectional driving scheme has been designed. A continuous wheel motion is resolved through the Inverse Kinematics (IK) scheme, which generates robot motion compliant with the Non-Sliding Pure-Rolling (NSPR) condition. A higher-level scheme resolving a steering motion to comply with the non-holonomic constraint and to tackle the structural singularity is proposed. To improve the robot performance in presence to the unpredicted circumstances, the IK scheme has been enhanced with the introduction of a new reactive support polygon adaptation task. To this end, a novel quadratic programming task has been designed to push the system Support Polygon Vertices (SPVs) away from the robot Centre of Mass (CoM), while respecting the leg workspace limits. The proposed task has been expressed through the developed SPV model to account for the hardware limits. The omnidirectional driving and reactive control schemes have been verified in the simulation and hardware experiments. To that end, the simulator for the CENTAURO robot that models the actuation dynamics and the software framework for the locomotion research have been developed.