The investigators will study how to enable biped robots to execute truly dynamic movements such as running, jumping, turning, starting rapidly, and stopping suddenly. For a biped to react flexibly and intelligently to changing surroundings, it must be endowed with a greater set of functional behaviors than those that are currently realized. The project will address four fundamental challenges common to periodic and aperiodic dynamic movements. First, analytical modeling and high fidelity dynamic simulation will be combined to develop quantitative metrics of dynamic stability based upon the level of intelligence required for recovery from disturbance. Second, evolutionary robotics strategies will be employed to efficiently search the large, discontinuous, and irregular solution space in order to generate truly dynamic movements. Third, a high-speed, drift-resistant sensing approach will be developed that is robust to large impacts with the environment by dividing the localization and attitude sensing problem into stance and flight segments. Fourth, an efficient, hybrid actuation approach will be developed in which electric motors actuate the hip degrees of freedom directly and, together with a novel air-spring system, provide the axial thrust. Finally, the work addressing these four challenges will be experimentally validated by constructing a three-dimensional biped capable of robustly executing fundamental dynamic movements. The results of this project will broadly impact the field of legged robotics since the core issues of dynamic stability, complex motion generation, high-speed sensing, and efficient actuation are not unique to a bipedal structure.
