Stable bipedal locomotion on granular terrain

The domain of legged robotics is a significant focus within the field of robotics, owing to their versatile utility across a spectrum of dynamically varied environments. Despite extensive research, the exploration of legged locomotion on granular and pliable terrains remains underdeveloped. This particular terrain type presents a considerable challenge to stability due to the propensity for foot sinkage and subsequent disturbances to the surrounding substrate, necessitating enhanced safety considerations in the trajectory planning and gait modification of robots. This thesis addresses the foot-terrain interaction and locomotion of bipedal robots traversing granular terrain. It introduces a theoretical framework based on Resistive Force Theory (RFT) for examining legged movement over such substrates, which are characterized by their malleable nature. The RFT model facilitates the quantification of interaction forces between the robot’s limbs and the terrain through a series of intrusion experiments. These experiments harness data from force plates and sophisticated three-dimensional kinematic constructs, which are critical in crafting the design of a stable and efficient gait. The investigation encompasses the impact of various factors, including foot morphology, gait trajectory, and the energetic considerations of power and work against velocity, as well as the yielding characteristics of the substrate. Empirical analysis is conducted through human locomotion experiments on deformable terrain. Here, data on joint positioning and movement patterns conducive to stable ambulation are amassed. The deployment of a force plate both beneath a sand walking test bed and on a solid surface permits the collection of metrics such as the center of pressure, center of mass, ground reaction forces, and joint moments. These ground reaction forces are then synthesized with the RFT framework to inform the design of gait trajectories. A suite of sophisticated sensors, including a Vicon motion capture system, an ATI Mini45 triaxial force/torque sensor, and force plates, are employed to gather data. Human walking is captured using a motion capture system and a comprehensive body model; this data is then digitized within a simulation platform to facilitate kinematic scrutiny and identification of pivotal gait characteristics on granular media. Experimental validation is conducted using a bespoke 6-degree-of-freedom bipedal robot. Subsequently, system identification techniques are employed to distill critical parameters for a bipedal robotic model. The ensuing control strategy is articulated through a low-level Proportional-Integral-Derivative (PID) control of each joint, calibrated to align with a Lagrangian dynamic model and harmonized with the RFT framework. Feedback linearization is utilized to orchestrate reference torques. The cornerstone of this research lies in the development of a mechatronic design for a bipedal robotic foot system and the formulation of a force modeling methodology predicated on human walking patterns. Additionally, a novel structural correction factor is proposed to address the specific issue of partial intrusion in compliant substrates. A central hypothesis tested is the variance in walking patterns between solid surfaces and granular media such as sand, with findings directly influencing the robotic model’s control design. The outcomes of this thesis have significant implications for the advancement of legged robotics on complex terrains, presenting a stride forward in the integration of human biomechanical insights into robotic design.