Transfer learning for reinforcement learning on a physical robot

By Samuel Barrett, Matthew E. Taylor, Peter Stone. 2010

80 citations recorded by the source corpus.

Subfields: Reinforcement Learning, Robotics & Control

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Reinforcement Learning

In Robotics & Control

Citations in the corpus

Learning Parameterized Skills (2012, methodology): “[3] introduced a similar method, in which the weights used to represent a Q-function are transferred to a target task.”
Sim-to-Real Robot Learning from Pixels with Progressive Nets (2016, background): “s. The paper also shows that the main failure point is the discrepancy in visual cues between simulation and reality. Partial success on transferring from simulation to a real robot has been reported [2, 7, 26]. They focus primarily on the problem of transfer from a more restricted simpler version o”
One-Shot Visual Imitation Learning via Meta-Learning (2017, methodology): “ demonstration of that task by using a dataset of demonstrations of many other tasks for meta-learning. Sharing information across tasks is by no means a new idea, e.g. by using task-to-task mappings [22], gating [23], and shared features [24]. These multi-task robotic learning methods consider the ”
Transfer from Simulation to Real World through Learning Deep Inverse Dynamics Model (2016, background): “There is a rich body of work focusing on adapting policies, rather than actions in the context of reinforcement learning [37] [1] [5].”
Meta Reinforcement Learning with Latent Variable Gaussian Processes (2018, background): “ith a single task. In a robotics context, solutions for multiple related tasks are often desired, e.g., for grasping multiple objects [22] or in robot games, such as robot table tennis [24] or soccer [3]. Much of the literature on meta and transfer learning in RL has focused on multi-task learning, ”
Transferring Autonomous Driving Knowledge on Simulated and Real Intersections (2017, background): “It is a well known problem that policies trained in simulation rarely work on real robots (Barrett et al., 2010).”
Multi-task policy search for robotics (2014, background): “, for grasping multiple objects [17] or in robot games, such as robot table tennis [18] or soccer [7].”
Learning legged swimming gaits from experience (2015, unknown): “This choice is inspired by authors such as Xu et al. [15], [16] who have recently shown that human guidance, in the forms of demonstrations or direct participation, are effective aids for learning practical robotic tasks.”
Reinforcement learning for non-prehensile manipulation: Transfer from simulation to physical system (2018, background): “one [31]. For sensitive and delicate systems, the only safe place to perform learning is in simulation. Transferring to real hardware can take many approaches as well, either through adaptation [32], [33], or incorporating uncertainty [34], [35]. This work focuses on using a physics simulator to tra”
Policy transfer via modularity and reward guiding (2017, background): “Furthermore, many transfer learning techniques require further training in the real-world to adapt the policy to the real-world dynamics [35].”
Generalizing Skills with Semi-Supervised Reinforcement Learning (2016, background): “A related but orthogonal problem is transfer learning (Taylor & Stone, 2009; Barrett et al., 2010), which attempts to use prior experience in one domain to improve training performance in another.”
Policy transfer via modularity (2017, background): “Furthermore, many transfer learning techniques require further training in the real-world to adapt the policy to the real-world dynamics [29].”