Desmurget, Michel and Claude Prablanc. Postural control of three-dimensional prehension movements. J. Neurophysiol. 77: 452–464, 1997. This experiment was carried out to test the hypothesis that three-dimensional upper limb movements could be initiated and controlled in the joint space via a mechanism comparing an estimate of the current postural state of the upper arm with a target value determined by one specific inverse static transform converting the coordinates of the object into a set of arm, forearm, and wrist angles. This hypothesis involves two main predictions: 1) despite joint redundancy, the posture reached by the upper limb should be invariant for a given context; and 2) a movement programmed in joint space should exhibit invariant characteristics of the joint covariation pattern as well as a corresponding variable hand path curvature in the task space. To test these predictions, we examined prehension movements toward a cylindrical object presented at a fixed spatial location and at various orientations without vision of the moving limb. Once presented, the object orientation was either kept constant (unperturbed trials) or suddenly modified at movement onset (perturbed trials). Three-dimensional movement trajectories were analyzed in both joint and task spaces. For the unperturbed trials, the task space analysis showed a variable hand path curvature depending on object orientation. The joint space analysis showed that the seven final angles characterizing the upper limb posture at hand-to-object contact varied monotonically with object orientation. At a dynamic level, movement onset and end were nearly identical for all joints. Moreover, for all joints having a monotonic variation, maximum velocity occurred almost simultaneously. For the elbow, the only joint presenting a reversal, the reversal was synchronized with the time to peak velocity of the other joint angles. For the perturbed trials, a smooth and complete compensation of the movement trajectory was observed in the task space. At a static level the upper limb final posture was identical to that obtained when the object was initially presented at the orientation following the perturbation. This result was particularly remarkable considering the large set of comfortable postures allowed by joint redundancy. At a dynamic level, the joints' covariation pattern was updated to reach the new target posture. The initial synergies were not disrupted by the perturbation, but smoothly modified, the different joints' movements ending nearly at the same time. Taken together, these results support the hypothesis that prehension movements are initiated and controlled in the joint space on the basis of a joint angular error vector rather than a spatial error vector.
A probabilistic Programming by Demonstration framework handling constraints in joint space and task space
