Robust and Efficient Quadrupedal Locomotion: The Role of Compliance

Quadrupeds have distinct advantages over humanoid and wheeled robots. They potentially possess superior mobility and dexterity in unstructured and challenging environments, which make them ideal candidates for inspection and exploration. However, despite recent advances, several challenges still undermine the complete deployability of these robots in harsh outdoor scenarios. Animals, instead, excel in locomoting in real-world scenarios with seeming ease and agility. In vertebrates, this is enabled mainly by the complex interactions among the musculoskeletal, sensory, and neural control systems. However, recent studies highlight the role of local physical loops, favored by body compliance and mechanics, that have very little to do with the central nervous system. Indeed, animals hinge on and exploit the intelligence embodied in their soft springy tissues, such as muscles and tendons, to robustly adapt to the features of the environment. Compliance can be modulated by co-contraction of muscles and enables high energy efficiency by periodically storing and releasing energy. The emerging soft and compliant robotics paradigm strives to pursue these principles. Nevertheless, this trend does not find a match in quadrupedal robots. Currently, strong simplifying assumptions and template models - e.g., point contact hypothesis or linear inverted pendulum model - deprive state-of-the-art locomotion methods of vital information about the mechanical properties and system compliance. Indeed, a general lack of attention towards more sophistication prevails in the three domains of planning, control, and design. In this context, the present dissertation discusses the role that compliance can play in quadrupedal robotics to overcome challenges related to the uncertainties of real-world unstructured scenarios. Some novel and unexplored solutions are introduced. These leverage compliance to improve the robustness and efficiency of quadrupedal locomotion by addressing three relevant shortcomings of current platforms: the inadequacy of motion planners in providing efficient trajectories for soft articulated quadrupeds, the absence of general frameworks for varying the mechanical impedance through control, and the lack of sophistication in mechanical foot design.