Pushing Fast and Slow: Task-Adaptive MPC for Pushing Manipulation Under Uncertainty
share
We propose a model predictive control approach to pushing based manipulation. The key feature of the algorithm is that it can adapt to the accuracy requirements of a task, by slowing down and generating "careful" motion when the task requires high accuracy, and by speeding up and moving fast when the task allows inaccuracy. We formulate the problem as an MDP and use an approximate online solution to the MDP. We use a trajectory optimizer with a deterministic model to suggest promising actions to the MDP, to reduce computation time spent on evaluating different actions. The trajectory optimizer is then initialized with trajectories with different speed profiles to generate a variety of actions for the MDP that can adapt to different tasks.
READ FULL TEXT