Planning and Control Strategies for Motion and Interaction of the Humanoid Robot COMAN+

FRANCESCO RUSCELLI
2021
Despite the majority of robotic platforms are still confined in controlled environments such as factories, thanks to the ever-increasing level of autonomy and the progress on human-robot interaction, robots are starting to be employed for different operations, expanding their focus from uniquely industrial to more diversified scenarios. Humanoid research seeks to obtain the versatility and dexterity of robots capable of mimicking human motion in any environment. With the aim of operating
more » ... -side with humans, they should be able to carry out complex tasks without posing a threat during operations. In this regard, locomotion, physical interaction with the environment and safety are three essential skills to develop for a biped. Concerning the higher behavioural level of a humanoid, this thesis addresses both ad-hoc movements generated for specific physical interaction tasks and cyclic movements for locomotion. While belonging to the same category and sharing some of the theoretical obstacles, these actions require different approaches: a general high-level task is composed of specific movements that depend on the environment and the nature of the task itself, while regular locomotion involves the generation of periodic trajectories of the limbs. Separate planning and control architectures targeting these aspects of biped motion are designed and developed both from a theoretical and a practical standpoint, demonstrating their efficacy on the new humanoid robot COMAN+, built at Istituto Italiano di Tecnologia. The problem of interaction has been tackled by mimicking the intrinsic elasticity of human muscles, integrating active compliant controllers. However, while state-of-the-art robots may be endowed with compliant architectures, not many can withstand potential system failures that could compromise the safety of a human interacting with the robot. This thesis proposes an implementation of such low-level controller that guarantees a fail-safe behaviour, removing the threat that a humanoid robot could pose if a [...]
doi:10.15167/ruscelli-francesco_phd2021-04-21 fatcat:2frgltov5zg5lhokwdkja3aoqa