Singulating an item from a pallet layer: Dual-arm manipulation with minimalistic end effectors by means of sampling-based MPC

Pavlos Theodosiadis, Alessandro Saccon, Thiago D. Simão

Keywords: robotic manipulation, sampling-based MPC, depalletization, dual-arm manipulation

Abstract:

This work addresses the challenge of picking an item from a orderly-arranged layer of objects by means of dual-arm manipulation with minimalistic end-effectors. The task is inspired by manual depalletization, a common material-handling process in logistics.
Successful execution of the task requires multiple sequential motions to isolate an object before both arms can hold it firmly. Motivated by the recent availability of parallel physics simulators, we explore the feasibility of sampling-based Model Predictive Control (MPC) to solve the combined motion planning and control problem online, as a complementary approach to offline-computed reinforcement learning policies. We propose a task-specific cost formulation and combine MPPI temperature adaptation with control input spline parametrization, to retain high-success rate with limited optimization parameters and thus reduced computational burden. We benchmark the effectiveness of the approach by means of a numerical study, against naive baseline implementations.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 9