Representing complex manipulation tasks, like scooping and pouring, as a sequence of constant screw motions in SE(3) allows us to extract the task-related constraints on the end-effector’s motion from kinesthetic demonstrations and transfer them to newer instances of the same tasks. This approach has been evaluated for complex manipulation tasks like scooping and pouring and also in the context of vertical containerised farming for transplanting and harvesting leafy crops. We have also developed an approach that allows us to transfer the task-related constraints between objects which are functionally similar but have different geometries. The notion of functional similarity is captured by a knowledge base. Using the screw-geometric structure of motion also allows us to generate motion plans for tasks which require object-environment contact like pivoting. More recently, we have developed a self-evaluation-based approach which allows the robot to compute the minimal set of kinesthetic demonstrations required to reliably perform tasks like pouring and scooping over a specified region of its workspace. </br>

Related Papers: