The ability to perform high-speed dynamic maneuvers is an important aspect of locomotion for bipedal animals such as humans. Running, jumping, and rapidly changing direction are fundamental dynamic maneuvers that contribute to the adaptability and performance required for bipeds to move through unstructured environments. A number of bipedal robots have been produced to investigate dynamic maneuvers. However, the level of performance demonstrated by biological systems has yet to be fully realized in a biped robot. One limiting factor in achieving comparable performance to animals is the lack of available control strategies that can successfully coordinate dynamic maneuvers. This thesis develops a control strategy for producing vertical jumping in a planar biped robot as a preliminary investigation into dynamic maneuvers. The control strategy was developed using a modular approach to allow adaptation to further dynamic maneuvers and robotic systems.
The control strategy was broken into two functional levels to separately solve the problems of planning and performing the jump maneuver. The jump is performed using a low-level controller, consisting of a state machine for determining the current phase of the jump and motor primitives for executing the joint motions required by the current phase. The motor primitives, described by open- and closed-loop control laws, were defined with numeric control parameters for modifying their performance. The high-level controller performs the task of planning the motion required to achieve the desired jump height. Fuzzy control, an intelligent control approach, was selected for the high-level controller. The fuzzy controller uses heuristic information about the biped system to select appropriate control parameters. This heuristic knowledge was implemented in a training algorithm. The training algorithm uses iterative jumps with error-based feedback to determine the control parameters to be implemented by the fuzzy controller.
The control strategy was developed and validated using a numerical simulation of the experimental biped KURMET. The simulation models the dynamics of the biped system and has demonstrated the ability of the control strategy to produce stable successive jumps with an approximate height of 0.575 m. The control strategy was also implemented on the experimental biped for a simplified case, resulting in stable successive jumps with a range of heights from 0.55 to 0.60 m.