| dc.description.abstract | There has been considerable research devoted to control and experimental realization of bipedal locomotion with the aim to develop a fully functional humanoid robot. Similar to humans, such a robot should be able to walk on a rough terrain, perform complex multitask operations, and also walk efficiently on level ground. Motivated with this future goal, this dissertation offers a control approach, a design solution, and an experimental demonstration of human-like actuated dynamic walking in biped robots.
During the conducted research, two preconditions to energy-efficient actuated dynamic walking have been identified. The first precondition, which is related to the control approach, precludes enforcing a predefined reference trajectory or other motion attributes of the walking cycle, such as stride length, stepping frequency or average forward speed, with high gain control. This has motivated the development of a control framework directly on the force level, that provides coordination without pre-specifying the motion of the system. The second precondition related to the robot design requires joint actuation which should not suppress passive joint motion (i.e., joints should be back-drivable like human joints). Utilization of back-drivable joint design allows the inertial motion to be exploited through walking rather than being suppressed by the actuation units. This has motivated the design of a 7-link biped robot with highly back-drivable joints, which is used in the experimental verification of the control framework.
The overall control philosophy is analytically developed, numerically investigated and experimentally realized. Instead of the usual robot walking characterized with stiff motion, bent-knee stance support, and no foot rotation, human-like dynamic walking is demonstrated here with (partially) ballistic swing, extended knee stance support and foot rotation during the (preemptive) ankle push-off phase. | |
