| Description |
This dissertation defines a new class of climbing robots, steering-plane bipeds, which encompasses a large number of existing climbing robots. Three major levels of motion planning are characterized which are common to this class of robots, namely, path planning, step planning, and gait planning. The unified presentation of related motion planning techniques is more generally applicable and more thorough than related algorithms in other literature, while more explicitly identifying limitations and tradeoffs due to alternate design choices within the class of steering-plane bipeds. A novel spline-based method for generating gaits is presented which uses separate path and time rate controls, and explicitly defined foot approach and departure directions that allows 1) a nominal guarantee of collision-free foot trajectories when close to the desired step configuration, 2) independent control of gait shape and speed, and 3) a unified representation of the four gait families of steering-plane bipeds: flipping, inchworm, step-through, and spinning gaits. This dissertation presents a thorough examination of the variations within each gait family, rather than merely presenting a representative instance of each. Concrete case studies applying the techniques of this dissertation are presented for optimizing the gaits for overall speed, energy efficiency, and minimum gripping force and moment. The results highlight that many common gaits in the literature are far from optimal. Results and general rules of thumb for gait planning are extracted that allow guidance for obtaining good results even if using alternate planning techniques without optimization. |
