We aim to build complex humanoid agents that integrate perception, motor control, and memory. In this work, we partly factor this problem into low-level motor control from proprioception and high-level coordination of the low-level skills informed by vision. We develop an architecture capable of surprisingly flexible, task-directed motor control of a relatively high-DoF humanoid body by combining pre-training of low-level motor controllers with a high-level, task-focused controller that switches among low-level sub-policies. The resulting system is able to control a physically-simulated humanoid body to solve tasks that require coupling visual perception from an unstabilized egocentric RGB camera during locomotion in the environment. For a supplementary video link, see https://youtu.be/7GISvfbykLE .

Sepsis is a dangerous condition that is a leading cause of patient mortality. Treating sepsis is highly challenging, because individual patients respond very differently to medical interventions and there is no universally agreed-upon treatment for sepsis. In this work, we explore the use of continuous state-space model-based reinforcement learning (RL) to discover high-quality treatment policies for sepsis patients. Our quantitative evaluation reveals that by blending the treatment strategy discovered with RL with what clinicians follow, we can obtain improved policies, potentially allowing for better medical treatment for sepsis.

In the field of Autonomous Driving, the system controlling the vehicle can be seen as an agent acting in a complex environment and thus naturally fits into the modern framework of Reinforcement Learning. However, learning to drive can be a challenging task and current results are often restricted to simplified driving environments. To advance the field, we present a method to adaptively restrict the action space of the agent according to its current driving situation and show that it can be used to swiftly learn to drive in a realistic environment based on the Deep Q-Network algorithm.

In hierarchical reinforcement learning a major challenge is determining appropriate low-level policies. We propose an unsupervised learning scheme, based on asymmetric self-play from Sukhbaatar et al. (2018), that automatically learns a good representation of sub-goals in the environment and a low-level policy that can execute them. A high-level policy can then direct the lower one by generating a sequence of continuous sub-goal vectors. We evaluate our model using Mazebase and Mujoco environments, including the challenging AntGather task. Visualizations of the sub-goal embeddings reveal a logical decomposition of tasks within the environment. Quantitatively, our approach obtains compelling performance gains over non-hierarchical approaches.

Policy gradient methods are widely used for control in reinforcement learning, particularly for the continuous action setting. There have been a host of theoretically sound algorithms proposed for the on-policy setting, due to the existence of the policy gradient theorem which provides a simplified form for the gradient. In off-policy learning, however, where the behaviour policy is not necessarily attempting to learn and follow the optimal policy for the given task, the existence of such a theorem has been elusive. In this work, we solve this open problem by providing the first off-policy policy gradient theorem. The key to the derivation is the use of $emphatic$ $weightings$. We develop a new actor-critic algorithm$\unicode{x2014}$called Actor Critic with Emphatic weightings (ACE)$\unicode{x2014}$that approximates the simplified gradients provided by the theorem. We demonstrate in a simple counterexample that previous off-policy policy gradient methods$\unicode{x2014}$particularly OffPAC and DPG$\unicode{x2014}$converge to the wrong solution whereas ACE finds the optimal solution.

Machine learning applications in medical imaging are frequently limited by the lack of quality labeled data. In this paper, we explore the self training method, a form of semi-supervised learning, to address the labeling burden. By integrating reinforcement learning, we were able to expand the application of self training to complex segmentation networks without any further human annotation. The proposed approach, reinforced self training (ReST), fine tunes a semantic segmentation networks by introducing a policy network that learns to generate pseudolabels. We incorporate an expert demonstration network, based on inverse reinforcement learning, to enhance clinical validity and convergence of the policy network. The model was tested on a pulmonary nodule segmentation task in chest X-rays and achieved the performance of a standard U-Net while using only 50% of the labeled data, by exploiting unlabeled data. When the same number of labeled data was used, a moderate to significant cross validation accuracy improvement was achieved depending on the absolute number of labels used.

Task-motion planning (TMP) addresses the problem of efficiently generating executable and low-cost task plans in a discrete space such that the (initially unknown) action costs are determined by motion plans in a corresponding continuous space. However, a task-motion plan can be sensitive to unexpected domain uncertainty and changes, leading to suboptimal behaviors or execution failures. In this paper, we propose a novel framework, TMP-RL, which is an integration of TMP and reinforcement learning (RL) from the execution experience, to solve the problem of robust task-motion planning in dynamic and uncertain domains. TMP-RL features two nested planning-learning loops. In the inner TMP loop, the robot generates a low-cost, feasible task-motion plan by iteratively planning in the discrete space and updating relevant action costs evaluated by the motion planner in continuous space. In the outer loop, the plan is executed, and the robot learns from the execution experience via model-free RL, to further improve its task-motion plans. RL in the outer loop is more accurate to the current domain but also more expensive, and using less costly task and motion planning leads to a jump-start for learning in the real world. Our approach is evaluated on a mobile service robot conducting navigation tasks in an office area. Results show that TMP-RL approach significantly improves adaptability and robustness (in comparison to TMP methods) and leads to rapid convergence (in comparison to task planning (TP)-RL methods). We also show that TMP-RL can reuse learned values to smoothly adapt to new scenarios during long-term deployments.

Eike Germann, Introduction to Reinforcement Learning What is Reinforcement Learning? How is it different from the machine learning we're familiar with? I'll present some foundational ideas (Markov decision processes, policy iteration, value iteration etc) and talk about their limitations. What algorithms are currently used to address those limitations and how do they do it? Based on these, I'll give a short overview of what RL is used currently used for - from training a machine to play space invaders to robotic movement in the real world.

We propose an exploration method that incorporates look-ahead search over basic learnt skills and their dynamics, and use it for reinforcement learning (RL) of manipulation policies . Our skills are multi-goal policies learned in isolation in simpler environments using existing multigoal RL formulations, analogous to options or macroactions. Coarse skill dynamics, i.e., the state transition caused by a (complete) skill execution, are learnt and are unrolled forward during lookahead search. Policy search benefits from temporal abstraction during exploration, though itself operates over low-level primitive actions, and thus the resulting policies does not suffer from suboptimality and inflexibility caused by coarse skill chaining. We show that the proposed exploration strategy results in effective learning of complex manipulation policies faster than current state-of-the-art RL methods, and converges to better policies than methods that use options or parametrized skills as building blocks of the policy itself, as opposed to guiding exploration. We show that the proposed exploration strategy results in effective learning of complex manipulation policies faster than current state-of-the-art RL methods, and converges to better policies than methods that use options or parameterized skills as building blocks of the policy itself, as opposed to guiding exploration.

We consider the problem of high-level strategy selection in the adversarial setting of real-time strategy games from a reinforcement learning perspective, where taking an action corresponds to switching to the respective strategy. Here, a good strategy successfully counters the opponent's current and possible future strategies which can only be estimated using partial observations. We investigate whether we can utilize the full game state information during training time (in the form of an auxiliary prediction task) to increase performance. Experiments carried out within a StarCraft: Brood War bot against strong community bots show substantial win rate improvements over a fixed-strategy baseline and encouraging results when learning with the auxiliary task.