Considerations for and Implementations of Deliberative and Reactive Motion Planning Strategies for the Novel Actuated Rimless Spoke Wheel Robot IMPASS for the Two-Dimensional Sagittal Plane
IMPASS is a novel spoke-wheel robot invented by researchers at the Robotics and Mechanisms Lab (RoMeLa) at Virginia Tech. The robot is driven by a rimless spoke wheel which can alter the length of any given spoke in the hub. This form of novel locomotion combines the efficiency of a wheeled robot and the mobility of a legged robot, arriving at a very practical mobility platform. A highly mobile robot such as IMPASS could prove very valuable in applications where the terrain is complex and dangerous, such as search and rescue, reconnaissance, or anti-terror response. A prototype has been constructed that effectively demonstrates the actuated spoke wheel concept using two wheels containing six spokes each.
Manually controlling the motion of two wheels and twelve spokes would be a daunting task for any operator. Due to this inherent complexity, automated motion control is a necessity for the IMPASS platform. The work presented here will discuss two different approaches to the motion planning problem for the two-dimensional sagittal plane. The first approach is deliberative in nature and depends on fairly accurate terrain sensing. The motion planning first decides on a set of contact points based on obstacle configurations and a Lagrangian interpolation of the terrain. A lower level motion planning component then executes the movements that guide the spoke ends to the contact points. The second motion planning approach is reactive in nature. Proprioceptive and tactile sensors are used to determine the robot's pose and immediate surroundings. These sensors directly affect the motion profile of the robot. The reactive approach follows much simpler logic, which theoretically will make it more robust.
Motion planning strategies were tested in simulation and on the IMPASS prototype. Both strategies proved to be well suited for different applications. The deliberative control was very successful in a structured environment, whereas the reactive control was able to cross a wider variety of terrain. The results from the testing also provided some insight into variables introduced by the hardware. Future improvements to the motion planning control include accounting for these variables in the hardware and eventually developing three-dimensional motion planning algorithms based on the lessons learned from the two-dimension case.