Thesis title: Generation and Control of Motion for 3D Humanoids and Steerable WMRs
Mobile robots have the potentiality to accomplish complex tasks because of their capability of adapting to different scenarios. This is particularly true for humanoid robots, which can navigate unstructured environments thanks to their anthropomorphic structure, and steerable wheeled mobile robots (WMRs), which present an increased mobility with respect to their non-omnidirectional counterpart.
In this thesis, we focus on the problem of generation and control of motion for humanoid robots in 3D, and steerable WMRs. In particular, when developing a framework for locomotion, one should take into account how the robot perceives its surroundings, how to move, and how to actually perform the motion. These problems, which lie under the categories of perception, motion planning, and control, must be solved efficiently and simultaneously.
In the first part of this manuscript, we study the problem of motion generation for humanoid robots in a world of stairs, a particular kind of uneven terrain where all contact surfaces are piecewise-horizontal. We present a framework composed of a RRT*-based footstep planner, a gait generation scheme based on Model Predictive Control (MPC), a mapping module, and a localization module. The footstep planner, which plans a feasible sequence of footsteps using an elevation map, exploits the time the robot takes to complete a step to replan the footsteps, improving the plan and taking into account changes in the map, which may occur due to the presence of dynamic obstacles. In order to improve the reliability of locomotion, external disturbances and pushes must be considered. To address these scenarios, we present Feasibility-Aware Plan Adaptation (FAPA), a module for adapting footstep plans (positions, orientations, and timings) in such a way to guarantee the feasibility of the subsequence MPC stage. FAPA allows to sustain external pushes on stairs, allowing the humanoid to safely complete locomotion tasks.
In the second part of this manuscript, we study the problem of motion control for steerable WMRs (SWMRs). The development of control schemes for this kind of platform is not trivial, due to the presence of kinematic singularities, which must be taken into account in order to properly make the robot move. We propose a framework for trajectory tracking of SWMRs using Nonlinear MPC (NMPC) based on the real-time iteration scheme. The NMPC generates feasible motions for the robot, taking into account both kinematic singularities of the mobile base, and bounds on driving and steering velocities. Our NMPC works alongside a finite state machine and an auxiliary trajectory generation scheme based on dynamic feedback linearization, which makes our framework capable of tracking trajectories without ever encountering singularities.
The proposed methods are validated in simulation using CoppeliaSim, and in experimental settings using the Neobotix MPO-700 platform.