Contact-Based Sensing, Modeling and Control for Smaller-Sized Humanoid and Multi-Legged Robots

This dissertation addresses issues in two large classes of robotic systems: humanoid and multi-legged systems. Such robots interact with the physical environment as a part of their normal function when executing real-world applications. Many capable robotic systems are large and costly. However, contact-based sensing, modeling and control remain challenges to many smaller-sized humanoid robots in object manipulation and to smaller multi-legged robots in complex terrain traversal. Compared to larger-sized humanoid robots with powerful actuators, smaller humanoids are safer and more affordable for in-home applications. However, their manipulation capability is limited because they lack accurate sensing devices and associated planning and control algorithms. Likewise, many critical applications require smaller multi-legged robots to move through cluttered environments such as rubble in search and rescue scenarios. Specifically, the complex terrain structure and the large perturbations in the robot's movement invalidate traditional path planning methods using obstacle avoidance. This dissertation therefore aims to expand the capacity of inexpensive smaller-sized robots. For humanoid robots used in manipulation tasks, lightweight, low-cost and accurate sensing devices are designed and associated planning and control algorithms are developed, which enable the robot to safely execute manipulation tasks. For multi-legged systems, this dissertation studies contact physics in the dynamic interactions between the robots and large obstacles using experiments, physics-based and mathematical modelings. The resulting capabilities developed here expand the usefulness of these inexpensive smaller-sized robots