Addressing Tasks Through Robot Adaptation

Developing flexible, broadly capable systems is essential for robots to move out of factories and into our daily lives, functioning as responsive agents that can handle whatever the world throws at them. This dissertation focuses on two kinds of robot adaptation. Modular self-reconfigurable robots (MSRR) adapt to the requirements of their task and environments by transforming themselves. By rearranging the connective structure of their component robot modules, these systems can assume different morphologies: for example, a cluster of modules might configure themselves into a car to maneuver on flat ground, a snake to climb stairs, or an arm to pick and place objects. Conversely, environment augmentation is a strategy in which the robot transforms its environment to meet its own needs, adding physical structures that allow it to overcome obstacles. In both areas, the presented work includes elements of hardware design, algorithms, and integrated systems, with the common goal of establishing these methods of adaptation as viable strategies to address tasks. The research takes a systems-level view of robotics, placing particular emphasis on experimental validation in hardware.