DGBench: An Open-Source, Reproducible Benchmark for Dynamic Grasping

Abstract

We present a robot base placement and control method that enables a mobile manipulator to gracefully recover from manipulation failures while performing tasks on-the-move. A mobile manipulator in motion has a limited window to complete a task, unlike when stationary where it can make repeated attempts until successful. Existing approaches to manipulation on-the-move are typically based on open-loop execution of planned trajectories which does not allow the base controller to react to manipulation failures, slowing down or stopping as required. To overcome this limitation, we present a reactive base control method that repeatedly evaluates the best base placement given the robot's current state, the immediate manipulation task, as well as the next part of a multi-step task. The result is a system that retains the reliability of traditional mobile manipulation approaches where the base comes to a stop, but leverages the performance gains available by performing manipulation on-the-move. The controller keeps the base in range of the target for as long as required to recover from manipulation failures while making as much progress as possible toward the next objective. See benburgesslimerick. github.io/MotM-FailureRecovery for videos of experiments.

An Architecture for Reactive Mobile Manipulation On-The-Move

“…3) Failure Recovery: The base control system presented in this work reactively adjusts the base motion to keep the object in reach until the grasp is completed. This is more thoroughly investigated in our related work presented in [28]. However, we have not implemented a robust method for detecting grasp failures in the real world.…”

“…We presented a base control method that integrates with the architecture presented in [1] to enable reactive manipulation on-the-move in complex environments with dynamic obstacles. Further, the QP in the redundancy resolution module is augmented to include an inequality constraint that provides collision avoidance for the manipulator.…”

“…Obstacle avoidance for the manipulator is enabled by modifying the redundancy resolution controller described in [1]. This controller solves a Quadratic Program (QP) to calculate joint velocities for a given desired end-effector and base velocity.…”

“…This is particularly valuable in multi-step tasks where, for example, the robot can be driving towards a second location while picking up an object. Recent works have explored control methods for achieving such Manipulation on-the-Move (MotM) [1], [2].…”

“…In our previous work [1], we introduced an architecture for achieving reactive MotM and demonstrated grasping of unpredictable, dynamic objects while on the move. However, the implementation presented previously uses a mobile base controller that does not consider obstacle avoidance.…”