Distributionally Robust Optimization (DRO) has enabled to prove the equivalence between robustness and regularization in classification and regression, thus providing an analytical reason why regularization generalizes well in statistical learning. Although DRO's extension to sequential decision-making overcomes $\textit{external uncertainty}$ through the robust Markov Decision Process (MDP) setting, the resulting formulation is hard to solve, especially on large domains. On the other hand, existing regularization methods in reinforcement learning only address $\textit{internal uncertainty}$ due to stochasticity. Our study aims to facilitate robust reinforcement learning by establishing a dual relation between robust MDPs and regularization. We introduce Wasserstein distributionally robust MDPs and prove that they hold out-of-sample performance guarantees. Then, we introduce a new regularizer for empirical value functions and show that it lower bounds the Wasserstein distributionally robust value function. We extend the result to linear value function approximation for large state spaces. Our approach provides an alternative formulation of robustness with guaranteed finite-sample performance. Moreover, it suggests using regularization as a practical tool for dealing with $\textit{external uncertainty}$ in reinforcement learning methods.

When learning policies for robot control, the real-world data required is typically prohibitively expensive to acquire, so learning in simulation is a popular strategy. Unfortunately, such polices are often not transferable to the real world due to a mismatch between the simulation and reality, called 'reality gap'. Domain randomization methods tackle this problem by randomizing the physics simulator (source domain) according to a distribution over domain parameters during training in order to obtain more robust policies that are able to overcome the reality gap. Most domain randomization approaches sample the domain parameters from a fixed distribution. This solution is suboptimal in the context of sim-to-real transferability, since it yields policies that have been trained without explicitly optimizing for the reward on the real system (target domain). Additionally, a fixed distribution assumes there is prior knowledge about the uncertainty over the domain parameters. Thus, we propose Bayesian Domain Randomization (BayRn), a black box sim-to-real algorithm that solves tasks efficiently by adapting the domain parameter distribution during learning by sampling the real-world target domain. BayRn utilizes Bayesian optimization to search the space of source domain distribution parameters which produce a policy that maximizes the real-word objective, allowing for adaptive distributions during policy optimization. We experimentally validate the proposed approach by comparing against two baseline methods on a nonlinear under-actuated swing-up task. Our results show that BayRn is capable to perform direct sim-to-real transfer, while significantly reducing the required prior knowledge.

We study the exploration problem with approximate linear action-value functions in episodic reinforcement learning under the notion of low inherent Bellman error, a condition normally employed to show convergence of approximate value iteration. First we relate this condition to other common frameworks and show that it is strictly more general than the low rank (or linear) MDP assumption of prior work. Second we provide an algorithm with a high probability regret bound $\widetilde O(\sum_{t=1}^H d_t \sqrt{K} + \sum_{t=1}^H \sqrt{d_t} \IBE K)$ where $H$ is the horizon, $K$ is the number of episodes, $\IBE$ is the value if the inherent Bellman error and $d_t$ is the feature dimension at timestep $t$. In addition, we show that the result is unimprovable beyond constants and logs by showing a matching lower bound. This has two important consequences: 1) the algorithm has the optimal statistical rate for this setting which is more general than prior work on low-rank MDPs 2) the lack of closedness (measured by the inherent Bellman error) is only amplified by $\sqrt{d_t}$ despite working in the online setting. Finally, the algorithm reduces to the celebrated \textsc{LinUCB} when $H=1$ but with a different choice of the exploration parameter that allows handling misspecified contextual linear bandits. While computational tractability questions remain open for the MDP setting, this enriches the class of MDPs with a linear representation for the action-value function where statistically efficient reinforcement learning is possible.

In cooperative multi-agent reinforcement learning (MARL), how to design a suitable reward signal to accelerate learning and stabilize convergence is a critical problem. The global reward signal assigns the same global reward to all agents without distinguishing their contributions, while the local reward signal provides different local rewards to each agent based solely on individual behavior. Both of the two reward assignment approaches have some shortcomings: the former might encourage lazy agents, while the latter might produce selfish agents. In this paper, we study reward design problem in cooperative MARL based on packet routing environments. Firstly, we show that the above two reward signals are prone to produce suboptimal policies. Then, inspired by some observations and considerations, we design some mixed reward signals, which are off-the-shelf to learn better policies. Finally, we turn the mixed reward signals into the adaptive counterparts, which achieve best results in our experiments. Other reward signals are also discussed in this paper. As reward design is a very fundamental problem in RL and especially in MARL, we hope that MARL researchers can rethink the rewards used in their systems.

Internet of things (IoT) devices are becoming increasingly popular thanks to many new services and applications they offer. However, in addition to their many benefits, they raise privacy concerns since they share fine-grained time-series user data with untrusted third parties. In this work, we study the privacy-utility trade-off (PUT) in time-series data sharing. Existing approaches to PUT mainly focus on a single data point; however, temporal correlations in time-series data introduce new challenges. Methods that preserve the privacy for the current time may leak significant amount of information at the trace level as the adversary can exploit temporal correlations in a trace. We consider sharing the distorted version of a user's true data sequence with an untrusted third party. We measure the privacy leakage by the mutual information between the user's true data sequence and shared version. We consider both instantaneous and average distortion between the two sequences, under a given distortion measure, as the utility loss metric. To tackle the history-dependent mutual information minimization, we reformulate the problem as a Markov decision process (MDP), and solve it using asynchronous actor-critic deep reinforcement learning (RL). We apply our optimal data release policies to location trace privacy scenario, and evaluate the performance of the proposed policy numerically.

We present a novel technique called Dynamic Experience Replay (DER) that allows Reinforcement Learning (RL) algorithms to use experience replay samples not only from human demonstrations but also successful transitions generated by RL agents during training and therefore improve training efficiency. It can be combined with an arbitrary off-policy RL algorithm, such as DDPG or DQN, and their distributed versions. We build upon Ape-X DDPG and demonstrate our approach on robotic tight-fitting joint assembly tasks, based on force/torque and Cartesian pose observations. In particular, we run experiments on two different tasks: peg-in-hole and lap-joint. In each case, we compare different replay buffer structures and how DER affects them. Our ablation studies show that Dynamic Experience Replay is a crucial ingredient that either largely shortens the training time in these challenging environments or solves the tasks that the vanilla Ape-X DDPG cannot solve. We also show that our policies learned purely in simulation can be deployed successfully on the real robot. The video presenting our experiments is available at https://sites.google.com/site/dynamicexperiencereplay

Deep reinforcement learning (RL) algorithms have recently achieved remarkable successes in various sequential decision making tasks, leveraging advances in methods for training large deep networks. However, these methods usually require large amounts of training data, which is often a big problem for real-world applications. One natural question to ask is whether learning good representations for states and using larger networks helps in learning better policies. In this paper, we try to study if increasing input dimensionality helps improve performance and sample efficiency of model-free deep RL algorithms. To do so, we propose an online feature extractor network (OFENet) that uses neural nets to produce good representations to be used as inputs to deep RL algorithms. Even though the high dimensionality of input is usually supposed to make learning of RL agents more difficult, we show that the RL agents in fact learn more efficiently with the high-dimensional representation than with the lower-dimensional state observations. We believe that stronger feature propagation together with larger networks (and thus larger search space) allows RL agents to learn more complex functions of states and thus improves the sample efficiency. Through numerical experiments, we show that the proposed method outperforms several other state-of-the-art algorithms in terms of both sample efficiency and performance.

In this article, we study a Radio Resource Allocation (RRA) that was formulated as a non-convex optimization problem whose main aim is to maximize the spectral efficiency subject to satisfaction guarantees in multiservice wireless systems. This problem has already been previously investigated in the literature and efficient heuristics have been proposed. However, in order to assess the performance of Machine Learning (ML) algorithms when solving optimization problems in the context of RRA, we revisit that problem and propose a solution based on a Reinforcement Learning (RL) framework. Specifically, a distributed optimization method based on multi-agent deep RL is developed, where each agent makes its decisions to find a policy by interacting with the local environment, until reaching convergence. Thus, this article focuses on an application of RL and our main proposal consists in a new deep RL based approach to jointly deal with RRA, satisfaction guarantees and Quality of Service (QoS) constraints in multiservice celular networks. Lastly, through computational simulations we compare the state-of-art solutions of the literature with our proposal and we show a near optimal performance of the latter in terms of throughput and outage rate.

In this paper, we consider the problem of building learning agents that can efficiently learn to navigate in constrained environments. The main goal is to design agents that can efficiently learn to understand and generalize to different environments using high-dimensional inputs (a 2D map), while following feasible paths that avoid obstacles in obstacle-cluttered environment. To achieve this, we make use of traditional path planning algorithms, supervised learning, and reinforcement learning algorithms in a synergistic way. The key idea is to decouple the navigation problem into planning and control, the former of which is achieved by supervised learning whereas the latter is done by reinforcement learning. Specifically, we train a deep convolutional network that can predict collision-free paths based on a map of the environment-- this is then used by a reinforcement learning algorithm to learn to closely follow the path. This allows the trained agent to achieve good generalization while learning faster. We test our proposed method in the recently proposed Safety Gym suite that allows testing of safety-constraints during training of learning agents. We compare our proposed method with existing work and show that our method consistently improves the sample efficiency and generalization capability to novel environments.

The widescale deployment of Autonomous Vehicles (AV) seems to be imminent despite many safety challenges that are yet to be resolved. It is well known that there are no universally agreed Verification and Validation (VV) methodologies to guarantee absolute safety, which is crucial for the acceptance of this technology. Existing standards focus on deterministic processes where the validation requires only a set of test cases that cover the requirements. Modern autonomous vehicles will undoubtedly include machine learning and probabilistic techniques that require a much more comprehensive testing regime due to the non-deterministic nature of the operating design domain. A rigourous statistical validation process is an essential component required to address this challenge. Most research in this area focuses on evaluating system performance in large scale real-world data gathering exercises (number of miles travelled), or randomised test scenarios in simulation. This paper presents a new approach to compute the statistical characteristics of a system's behaviour by biasing automatically generated test cases towards the worst case scenarios, identifying potential unsafe edge cases.We use reinforcement learning (RL) to learn the behaviours of simulated actors that cause unsafe behaviour measured by the well established RSS safety metric. We demonstrate that by using the method we can more efficiently validate a system using a smaller number of test cases by focusing the simulation towards the worst case scenario, generating edge cases that correspond to unsafe situations.