The goal of offline reinforcement learning is to learn a policy from a fixed dataset, without further interactions with the environment. This setting will be an increasingly more important paradigm for real-world applications of reinforcement learning such as robotics, in which data collection is slow and potentially dangerous. Existing off-policy algorithms have limited performance on static datasets due to extrapolation errors from out-of-distribution actions. This leads to the challenge of constraining the policy to select actions within the support of the dataset during training. We propose to simply learn the Policy in the Latent Action Space (PLAS) such that this requirement is naturally satisfied. We evaluate our method on continuous control benchmarks in simulation and a deformable object manipulation task with a physical robot. We demonstrate that our method provides competitive performance consistently across various continuous control tasks and different types of datasets, outperforming existing offline reinforcement learning methods with explicit constraints

In previous works, tactile sensing has been mainly used in manipulation to get some information about the material of the object that is being manipulated, in order to, for example, prevent slip and improve the quality of the grasp. Humans can leverage tactile information for other dexterous manipulation tasks like manipulating an object without relying on vision using sliding. Comparing to different modalities of perception, tactile sensing has some advantages over vision. It provides location and geometry information of the cloth portion being grasped, which would be otherwise occluded when relying on vision only. Tactile sensors can provide undisturbed information about the status of the current grasp. Possible applications of learning skills such as sliding are: cloth folding, rope manipulation, etc.

We are using a sawyer equipped with a WSG 32 gripper w/ WSG DSA tactile sensor fingers. The task that we are trying to do is this: One point of the cloth is fixed, the gripper grasps the cloth edge near this point and slides to get to the end of the cloth. The policy is trained on real robot using TD3.

Some robots can interact with humans using natural language, and identify service requests through human-robot dialog. We develop a dialog agent for robots that is able to interpret user commands using a semantic parser, while asking clarification questions using a probabilistic dialog manager. This dialog agent is able to augment its language capabilities by adding new entities and learning new ways of referring to existing entities. We extensively evaluated our dialog system in simulation as well as with human participants through MTurk and real-robot platforms.