DL and DRL for Intelligent Manipulation

We study ways to close the loop between motion generation and decision making through the use of deep learning and deep reinforcement learning. In recent years, DRL has seen an explosion in activity in Deep Q-Learning, (Determinisitc) Policy Gradients, Exploration, Hierarchical RL, Intrinsic Rewards, Imitation Learning, and other areas.

We leverage the techniques in dexterous manipulation as well as human robot collaboration and apply it in areas of industry and service.

Related Work:

• Efficiency and Tuning
  • PlanQ: Planning with Q-values in Sparse
    Rewards.
  • Hyperparameter Auto-tuning in Self-Supervised Robotic Learning
  • Invariant Transform Experience Replay
• Skills
  • Learning to Rock-and-Walk:
    Dynamic, Non-Prehensile, and Underactuated Object Locomotion through Reinforcement
    Learning.
  • Learning to Pick by Digging: Data-
    Driven Dig-Grasping for Bin Picking from Clutter.
  • Expert Demonstrations
    in Robot Cooperation with Multi-Agent Reinforcement Learning.
  • Towards Safe Control of Continuum Manipulator Using Shielded Multiagent Reinforcement Learning