AIMS CDT Project Report : Motion Planning To Smoothly Intercept Moving Objects

Kinematic planning techniques are well researched and have been shown to be very effective in scenarios with static objects. However, these approaches fail when interacting with dynamic environments where objects are moving. Among the key challenges faced in picking up moving objects are that planning needs to account for firstly whether the solution will intercept the object at an appropriate time, and secondly obey the constraint that we typically don’t want a large deceleration, in order to avoid breakages and spillages. In this research we build on the search-based kinodynamic motion planning algorithm presented by Menon et al. which generates a time-parameterised trajectory for whole-body motion in order to intercept an object at the earliest feasible point in it’s trajectory. Their research focussed only on planning in two spatial dimensions, working well in a conveyor belt environment at a fixed height. In this research we propose a fast, on-the-fly heuristic to counter the high dimensionality of the problem and expand the search space to plan in a 3D environment, greatly increasing the capabilities of the robot. We present the progress achieved so far and describe our plans for continuing this research. We verify our implementation in simulation on a 6-DOF environment, controlling the base rotation and 5-DOF of the arm of Toyota’s Human Support Robot (HSR) [1], to intercept a Cartesian goal coordinate at the earliest time.