A Task-Adaptive Deep Reinforcement Learning Framework for Dual-Arm Robot Manipulation

Closed-chain manipulation occurs when several robot arms perform tasks in cooperation. It is complex to control a dual-arm system because it requires flexible and adaptable operation ability to realize closed-chain manipulation. In this study, a deep reinforcement learning (DRL) framework based on actor-critic algorithm is proposed to drive the closed-chain manipulation of a dual-arm robotic system. The proposed framework is designed to train dual robot arms to transport a large object cooperatively. In order to sustain strict constraints of closed-chain manipulation, the actor part of the proposed framework is designed in a leader-follower mode. The leader part consists of a policy trained from the DRL algorithm and works on the leader arm. The follower part consists of an inverse kinematics solver based on Damped Least Squares (DLS) and works on the follower arm. Two experiments are designed to prove the task adaptability, one of which is manipulating an object to a random pose within a defined range, the other is manipulating a delicate structural object within a narrow space Note to Practitioners —In common industrial manipulation scenarios, there are requirements to employ robotic arms to transport a large object relative to the robotic arm, e.g., moving a payload onto a loader and assembling big craft parts. It is a cost-effective way to use a dual-arm system to extend the loading capacity of robotic arms while preserving the flexibility of manipulation. Moreover, the dual-arm system is expected to manipulate different objects without complicated reprogramming, especially in small batch production scenarios. This study proposes a task-adaptive deep reinforcement learning framework for dual-arm robot manipulation. The task adaptability includes two specific aspects, one being adaptability in targeting the pose, such as manipulating an object to a random pose within a specified range. The other is the adaptivity on the task prerequisites such as manipulating a delicate structural object within a narrow space. For future research, the dual-arm system may autonomously plan the grab positions, and additional investigations should address more common scenarios involving various object shapes.