We study continuous action reinforcement learning problems in which it is crucial that the agent interacts with the environment only through safe policies, i.e.,~policies that do not take the agent to undesirable situations. We formulate these problems as constrained Markov decision processes (CMDPs) and present safe policy optimization algorithms that are based on a Lyapunov approach to solve them. Our algorithms can use any standard policy gradient (PG) method, such as deep deterministic policy gradient (DDPG) or proximal policy optimization (PPO), to train a neural network policy, while guaranteeing near-constraint satisfaction for every policy update by projecting either the policy parameter or the action onto the set of feasible solutions induced by the state-dependent linearized Lyapunov constraints. Compared to the existing constrained PG algorithms, ours are more data efficient as they are able to utilize both on-policy and off-policy data. Moreover, our action-projection algorithm often leads to less conservative policy updates and allows for natural integration into an end-to-end PG training pipeline. We evaluate our algorithms and compare them with the state-of-the-art baselines on several simulated (MuJoCo) tasks, as well as a real-world indoor robot navigation problem, demonstrating their effectiveness in terms of balancing performance and constraint satisfaction. Videos of the experiments can be found in the following link: https://drive.google.com/file/d/1pzuzFqWIE710bE2U6DmS59AfRzqK2Kek/view?usp=sharing.

This is the first in a series of articles on reinforcement learning and OpenAI Gym. Suppose you're playing a video game. You enter a room with two doors. Behind Door 1 are 100 gold coins, followed by a passageway. Behind Door 2 is 1 gold coin, followed by a second passageway going in a different direction. Once you go through one of the doors, there is no going back.

Deep reinforcement learning provides a promising approach for vision-based control of real-world robots. However, the generalization of such models depends critically on the quantity and variety of data available for training. This data can be difficult to obtain for some types of robotic systems, such as fragile, small-scale quadrotors. Simulated rendering and physics can provide for much larger datasets, but such data is inherently of lower quality: many of the phenomena that make the real-world autonomous flight problem challenging, such as complex physics and air currents, are modeled poorly or not at all, and the systematic differences between simulation and the real world are typically impossible to eliminate. In this work, we investigate how data from both simulation and the real world can be combined in a hybrid deep reinforcement learning algorithm. Our method uses real-world data to learn about the dynamics of the system, and simulated data to learn a generalizable perception system that can enable the robot to avoid collisions using only a monocular camera. We demonstrate our approach on a real-world nano aerial vehicle collision avoidance task, showing that with only an hour of real-world data, the quadrotor can avoid collisions in new environments with various lighting conditions and geometry. Code, instructions for building the aerial vehicles, and videos of the experiments can be found at github.com/gkahn13/GtS

Deep Reinforcement Learning has been shown to be very successful in complex games, e.g. Atari or Go. These games have clearly defined rules, and hence allow simulation. In many practical applications, however, interactions with the environment are costly and a good simulator of the environment is not available. Further, as environments differ by application, the optimal inductive bias (architecture, hyperparameters, etc.) of a reinforcement agent depends on the application. In this work, we propose a multi-arm bandit framework that selects from a set of different reinforcement learning agents to choose the one with the best inductive bias. To alleviate the problem of sparse rewards, the reinforcement learning agents are augmented with surrogate rewards. This helps the bandit framework to select the best agents early, since these rewards are smoother and less sparse than the environment reward. The bandit has the double objective of maximizing the reward while the agents are learning and selecting the best agent after a finite number of learning steps. Our experimental results on standard environments show that the proposed framework is able to consistently select the optimal agent after a finite number of steps, while collecting more cumulative reward compared to selecting a sub-optimal architecture or uniformly alternating between different agents.

We address the challenge of effective exploration while maintaining good performance in policy gradient methods. As a solution, we propose diverse exploration (DE) via conjugate policies. DE learns and deploys a set of conjugate policies which can be conveniently generated as a byproduct of conjugate gradient descent. We provide both theoretical and empirical results showing the effectiveness of DE at achieving exploration, improving policy performance, and the advantage of DE over exploration by random policy perturbations.

In this paper, we investigate the use of global information to speed up the learning process and increase the cumulative rewards of multi-agent reinforcement learning (MARL) tasks. Within the actor-critic MARL, we introduce multiple cooperative critics from two levels of the hierarchy and propose a hierarchical critic-based MARL algorithm. In our approach, the agent is allowed to receive information from local and global critics in a competition task. The agent not only receives low-level details but also considers coordination from high levels to obtain global information for increasing operational performance. Here, we define multiple cooperative critics in a top-down hierarchy, called the Hierarchical Critic Assignment (HCA) framework. Our experiment, a two-player tennis competition task performed in the Unity environment, tested the HCA multi-agent framework based on the Asynchronous Advantage Actor-Critic (A3C) with Proximal Policy Optimization (PPO) algorithm. The results showed that the HCA framework outperforms the non-hierarchical critic baseline method on MARL tasks.

Reinforcement learning (RL) problems often feature deceptive local optima, and learning methods that optimize purely for reward signal often fail to learn strategies for overcoming them. Deep neuroevolution and novelty search have been proposed as effective alternatives to gradient-based methods for learning RL policies directly from pixels. In this paper, we introduce and evaluate the use of novelty search over agent action sequences by string edit metric distance as a means for promoting innovation. We also introduce a method for stagnation detection and population resampling inspired by recent developments in the RL community that uses the same mechanisms as novelty search to promote and develop innovative policies. Our methods extend a state-of-the-art method for deep neuroevolution using a simple-yet-effective genetic algorithm (GA) designed to efficiently learn deep RL policy network weights. Experiments using four games from the Atari 2600 benchmark were conducted. Results provide further evidence that GAs are competitive with gradient-based algorithms for deep RL. Results also demonstrate that novelty search over action sequences is an effective source of selection pressure that can be integrated into existing evolutionary algorithms for deep RL.

From 2017 to 2018 the number of scientific publications found via PubMed search using the keyword "Machine Learning" increased by 46% (4,317 to 6,307). The results of studies involving machine learning, artificial intelligence (AI), and big data have captured the attention of healthcare practitioners, healthcare managers, and the public at a time when Western medicine grapples with unmitigated cost increases and public demands for accountability. The complexity involved in healthcare applications of machine learning and the size of the associated data sets has afforded many researchers an uncontested opportunity to satisfy these demands with relatively little oversight. In a recent Nature Medicine article, "The Artificial Intelligence Clinician learns optimal treatment strategies for sepsis in intensive care," Komorowski and his coauthors propose methods to train an artificial intelligence clinician to treat sepsis patients with vasopressors and IV fluids. In this post, we will closely examine the claims laid out in this paper. In particular, we will study the individual treatment profiles suggested by their AI Clinician to gain insight into how their AI Clinician intends to treat patients on an individual level.

In many finite horizon episodic reinforcement learning (RL) settings, it is desirable to optimize for the undiscounted return - in settings like Atari, for instance, the goal is to collect the most points while staying alive in the long run. Yet, it may be difficult (or even intractable) mathematically to learn with this target. As such, temporal discounting is often applied to optimize over a shorter effective planning horizon. This comes at the cost of potentially biasing the optimization target away from the undiscounted goal. In settings where this bias is unacceptable - where the system must optimize for longer horizons at higher discounts - the target of the value function approximator may increase in variance leading to difficulties in learning. We present an extension of temporal difference (TD) learning, which we call TD($\Delta$), that breaks down a value function into a series of components based on the differences between value functions with smaller discount factors. The separation of a longer horizon value function into these components has useful properties in scalability and performance. We discuss these properties and show theoretic and empirical improvements over standard TD learning in certain settings.

Despite many algorithmic advances, our theoretical understanding of practical distributional reinforcement learning methods remains limited. One exception is Rowland et al. (2018)'s analysis of the C51 algorithm in terms of the Cram\'er distance, but their results only apply to the tabular setting and ignore C51's use of a softmax to produce normalized distributions. In this paper we adapt the Cram\'er distance to deal with arbitrary vectors. From it we derive a new distributional algorithm which is fully Cram\'er-based and can be combined to linear function approximation, with formal guarantees in the context of policy evaluation. In allowing the model's prediction to be any real vector, we lose the probabilistic interpretation behind the method, but otherwise maintain the appealing properties of distributional approaches. To the best of our knowledge, ours is the first proof of convergence of a distributional algorithm combined with function approximation. Perhaps surprisingly, our results provide evidence that Cram\'er-based distributional methods may perform worse than directly approximating the value function.