arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1802.10463cs.LGcs.AIcs.ROstat.ML
收藏

DiGrad:具有共享动作的多任务强化学习

DiGrad: Multi-Task Reinforcement Learning with Shared Actions

International Institute of Information Technology, Hyderabad(海得拉巴国际信息技术研究所);Indian Institute of Technology, Madras(马德拉斯印度理工学院)

Parijat Dewangan, S Phaniteja, K Madhava Krishna, Abhishek Sarkar, Balaraman Ravindran

摘要:

大多数强化学习算法对于复杂机器人系统中的多个任务的学习是低效的,其中不同的任务共享一组动作。在这样的环境中,可以利用共享的神经网络参数来学习复合策略,其同时执行多个任务。然而,这种复合策略可能会偏向于一个任务,或者来自不同任务的梯度相互抵消,使得学习不稳定,有时数据效率较低。在本文中,我们提出了一种新的方法,同时训练多个任务共享一组共同的行动在连续的行动空间,我们称之为DiGrad(差分策略梯度)。所提出的框架是基于差分策略梯度,可以容纳在一个单一的演员-评论家网络的多任务学习。我们还提出了一个简单的启发式的差分策略梯度更新,以进一步提高学习。在8连杆平面机械臂和27自由度仿人机器人上分别对3个和2个末端执行器的多目标可达性任务进行了学习测试。我们表明,我们的方法支持复杂机器人系统中的高效多任务学习,在连续动作空间中优于相关方法。

英文摘要:

Most reinforcement learning algorithms are inefficient for learning multiple tasks in complex robotic systems, where different tasks share a set of actions. In such environments a compound policy may be learnt with shared neural network parameters, which performs multiple tasks concurrently. However such compound policy may get biased towards a task or the gradients from different tasks negate each other, making the learning unstable and sometimes less data efficient. In this paper, we propose a new approach for simultaneous training of multiple tasks sharing a set of common actions in continuous action spaces, which we call as DiGrad (Differential Policy Gradient). The proposed framework is based on differential policy gradients and can accommodate multi-task learning in a single actor-critic network. We also propose a simple heuristic in the differential policy gradient update to further improve the learning. The proposed architecture was tested on 8 link planar manipulator and 27 degrees of freedom(DoF) Humanoid for learning multi-goal reachability tasks for 3 and 2 end effectors respectively. We show that our approach supports efficient multi-task learning in complex robotic systems, outperforming related methods in continuous action spaces.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑