Using Virtual Articulations to Operate High-DoF Inspection and Manipulation Motions

Abstract

We have developed a new operator interface system for high-DoF articulated robots based on the idea of allowing the operator to extend the robot's actual kinematics with virtual articulations. These virtual links and joints can model both primary task DoF and constraints on whole-robot coordinated motion. Unlike other methods, our approach can be applied to robots and tasks of arbitrary kinematic topology, and allows specifying motion with a scalable level of detail. We present hardware results where NASA/JPL's All-Terrain Hex-Legged Extra-Terrestrial Explorer (ATHLETE) executes previously challenging inspection and manipulation motions involving coordinated motion of all 36 of the robot's joints.

A graphical operations interface for modular surface systems

“…In [15] we also demonstrated the feasibility of our approach on the original ATHLETE hardware. We have yet to do physical experiments with the new modular TriATHLETE, which is little more than a few months old at the time of this writing.…”

“…This can be as straightforward and intuitive as clicking on any part of the model and dragging with the mouse, though there are also more quantitative ways to specify kinematic motions in our system. The system will enforce all constraints implied by the robot joints, and additional constraints on coordinated motion can also be defined using virtual joints and links, as we reported previously in [15]. Any subset of the model may also be locked at any time to limit the motion.…”

“…The operator simply specifies that the wheel forks must remained locked in space and then drags the pallet to specify motion. In general, any combination of links and joints can be locked or dragged simultaneously, and quantified motion ("tilt down 30 degrees") can also be defined using virtual joints and numeric trajectory input [15]. .…”

“…This is actually semantically meaningful: the kinematic constraints implied by a tree joint are always guaranteed to be maintained (and require no special computation), but the constraints implied by closure joints must be actively enforced by the system. We use an iterative least-squares numerical solver, a common technique in robotics [15,16]. This works well in general, but the iterative optimization style of constraint satisfaction means that (a) computation is continually expended to maintain closure joint constraints (in fact, the computational cost is typically…”

“…We have already presented our topology-independent forward/inverse kinematics computation and click-and-drag direct manipulation algorithms in an earlier publication [15], and do not focus on these aspects of the system here. Full details on both the topological and interactive direct manipulation algorithms are also given in [16].…”