Advanced Walking Skills for Bipedal Locomotion

Most control algorithms for current bipedal robots relied on the stability analysis of ZeroMoment Point (ZMP), which was rstly introduced by Vukobratović in 1968. However, control approaches based on ZMP have yet common shortcomings, such as, unnatural motions, high energy costs, high computational burden, absence of natural dynamics. Due to their disadvantages, many researchers studied the biological and biomechanical aspects of human locomotion, from which valuable ndings can be transferred to the design of control approaches for bipedal walking robots. Among them, a Bio-Inspired Behavior-Based Bipedal Locomotion Control (B4LC) was developed by Luksch based on the concepts found in human motion control with the hypothesis that a control system similar to human one will enable a robot to perform locomotion in a likewise fashion. Examining the achievement of the B4LC approach, it can be observed that several re nements and extensions are desirable. The current walking skills contain walking initiation, cyclic walking and standing with perturbations. Moreover, to extend the walking skills of a humanlike robot, certain advanced walking skills, such as, walking termination, curve walking, and upslope walking in walking are studied and control methods for them are designed based on biological and biomechanical literature. To enhance the robustness of bipedal walking in more challenging environments, a control structure is thus required. It features detection and rejection of unexpected disturbances during walking, especially the external pushes. Based on the processing of sensory information, di erent control strategies should be applied according to the state of the biped to come to a relative stable state. The external pushes can come from sagittal and frontal plane and happen during certain time instants of a gait cycle. Investigating the literature in push recovery of human walking leads to development of speci c control strategies for joints in lower limb. Besides the external disturbances, voluntary control from robot itself may also result in instability of the robot. Neuroscientists assume that ve electromyography (emg) components account for muscle activity during walking under di erent speeds. This nding brings us the implication for designing the corresponding motor patterns across di erent walking speeds. With approximating those ve emg components to Gaussian curves, it can deduce the problem as the search of parameters representing those curves, which are the mean values, the variances and the peak of amplitudes, respectively. By employing the Swarm Particle Optimization method, parameters for those motor patterns can be automatically found and optimized in terms of stability and energy cost. Analyzing all parameters within di erent walking speeds, construction of a function of parameters for motor patterns across walking speeds can be accomplished. Furthermore, the drawback of manually tuning process for parameters of the control units and of the biped model still persists. To nd optimal control parameters, the introduction of