Abstracts Computer Science

Advances in grasping and vehicle contact identification : analysis, design and testing of robust methods for underwater robot manipulation

by Edward R. Snow

This thesis focuses on improving the productivity of autonomous and telemanipulation systems consisting of a manipulator arm mounted to a free flying underwater vehicle. Part I minimizes system sensitivity to misalignment by developing a gripper and a suite of handles that passively self align when grasped. After presenting a gripper guaranteed to passively align cylinders we present several other self aligning handles. The mix of handle alignment and load resisting properties enables handles to be matched to the needs of each task. Part I concludes with a discussion of successful field use of the system on the Jason Remotely Operated Undersea Vehicle operated by the Woods Hole Oceanographic Institution. To enable the exploitation of contact with the environment to help stabilize the vehicle, Part II develops a technique which identifies the contact state of a planar vehicle interacting with a fixed environment. Knowing the vehicle geometry and velocity we identify kinematically feasible contact points, from which we construct the set of feasible contact models. The measured vehicle data violates each model’s constraints; we use the associated violation power and work to select the best overall model. Part II concludes with experimental confirmation of the contact identification techniques efficacy.