Posted by C. Daniel Freeman, Senior Software Engineer and Erik Frey, Staff Software Engineer, Google Research Reinforcement learning (RL)...

Object-centric world models provide structured representation of the scene and can be an important backbone in reinforcement learning and planning. However, existing approaches suffer in partially-observable environments due to the lack of belief states. In this paper, we propose Structured World Belief, a model for learning and inference of object-centric belief states. Inferred by Sequential Monte Carlo (SMC), our belief states provide multiple object-centric scene hypotheses. To synergize the benefits of SMC particles with object representations, we also propose a new object-centric dynamics model that considers the inductive bias of object permanence. This enables tracking of object states even when they are invisible for a long time. To further facilitate object tracking in this regime, we allow our model to attend flexibly to any spatial location in the image which was restricted in previous models. In experiments, we show that object-centric belief provides a more accurate and robust performance for filtering and generation. Furthermore, we show the efficacy of structured world belief in improving the performance of reinforcement learning, planning and supervised reasoning.

Pre-training Reinforcement Learning agents in a task-agnostic manner has shown promising results. However, previous works still struggle in learning and discovering meaningful skills in high-dimensional state-spaces, such as pixel-spaces. We approach the problem by leveraging unsupervised skill discovery and self-supervised learning of state representations. In our work, we learn a compact latent representation by making use of variational and contrastive techniques. We demonstrate that both enable RL agents to learn a set of basic navigation skills by maximizing an information theoretic objective. We assess our method in Minecraft 3D pixel maps with different complexities. Our results show that representations and conditioned policies learned from pixels are enough for toy examples, but do not scale to realistic and complex maps. To overcome these limitations, we explore alternative input observations such as the relative position of the agent along with the raw pixels.

Building loads consume roughly 40% of the energy produced in developed countries, a significant part of which is invested towards building temperature-control infrastructure. Therein, renewable resource-based microgrids offer a greener and cheaper alternative. This communication explores the possible co-design of microgrid power dispatch and building HVAC (heating, ventilation and air conditioning system) actuations with the objective of effective temperature control under minimised operating cost. For this, we attempt control designs with various levels of abstractions based on information available about microgrid and HVAC system models using the Deep Reinforcement Learning (DRL) technique. We provide control architectures that consider model information ranging from completely determined system models to systems with fully unknown parameter settings and illustrate the advantages of DRL for the design prescriptions.

To reduce the spread of COVID-19, and reduce hospitalizations and deaths resulting from infections, provinces and territories across Canada have imposed restrictions and limitations on the population based on Non-Pharmaceutical Intervention (NPI) strategies. One of the many factors that plays a role into the selection of NPIs is the use of epidemiological modelling. Using Reinforcement Learning, data scientists at Statistics Canada, in collaboration with partners at the Public Health Agency of Canada, helped determine the optimal set of population behaviours that minimize the spread of an infection within simulations to model and/or forecast the effect of a set of specific scenarios.

So far, most research on recommender systems focused on maintaining long-term user engagement and satisfaction, by promoting relevant and personalized content. However, it is still very challenging to evaluate the quality and the reliability of this content. In this paper, we propose FEBR (Expert-Based Recommendation Framework), an apprenticeship learning framework to assess the quality of the recommended content on online platforms. The framework exploits the demonstrated trajectories of an expert (assumed to be reliable) in a recommendation evaluation environment, to recover an unknown utility function. This function is used to learn an optimal policy describing the expert's behavior, which is then used in the framework to provide high-quality and personalized recommendations. We evaluate the performance of our solution through a user interest simulation environment (using RecSim). We simulate interactions under the aforementioned expert policy for videos recommendation, and compare its efficiency with standard recommendation methods. The results show that our approach provides a significant gain in terms of content quality, evaluated by experts and watched by users, while maintaining almost the same watch time as the baseline approaches.

Policy optimization methods are among the most widely-used methods in reinforcement learning. Its empirical success has been demonstrated in various domains such as computer games [Schulman et al., 2017] and robotics [Levine and Koltun, 2013]. However, due to its local-search nature, global optimality guarantees of policy optimization often rely on unrealistic assumptions to ensure global exploration (see e.g., [Abbasi-Yadkori et al., 2019, Agarwal et al., 2020b, Neu and Olkhovskaya, 2020, Wei et al., 2021]), making it theoretically less appealing compared to other methods. Motivated by this issue, a line of recent works [Cai et al., 2020, Shani et al., 2020, Agarwal et al., 2020a, Zanette et al., 2021] equip policy optimization with global exploration by adding exploration bonuses to the update, and prove favorable guarantees even without making extra exploratory assumptions. Moreover, they all demonstrate some robustness aspect of policy optimization (such as being able to handle adversarial losses or a certain degree of model misspecification). Despite these important progresses, however, many limitations still exist, including worse regret rates comparing to the best value-based or model-based approaches [Shani et al., 2020, Agarwal et al., 2020a, Zanette et al., 2021], or requiring full-information feedback on the entire loss function (as opposed to the more realistic bandit feedback) [Cai et al., 2020]. To address these issues, in this work, we propose a new type of exploration bonuses called dilated bonuses, which satisfies a certain dilated Bellman equation and provably leads to improved exploration compared to existing works (Section 3). We apply this general idea to advance the state-of-the-art of policy optimization for learning finite-horizon episodic MDPs with adversarial losses and bandit feedback. More specifically, our main results are: - First, in the tabular setting, addressing the main open question left in [Shani et al., 2020], we improve their Õ(T

Autonomous car racing is a challenging task in the robotic control area. Traditional modular methods require accurate mapping, localization and planning, which makes them computationally inefficient and sensitive to environmental changes. Recently, deep-learning-based end-to-end systems have shown promising results for autonomous driving/racing. However, they are commonly implemented by supervised imitation learning (IL), which suffers from the distribution mismatch problem, or by reinforcement learning (RL), which requires a huge amount of risky interaction data. In this work, we present a general deep imitative reinforcement learning approach (DIRL), which successfully achieves agile autonomous racing using visual inputs. The driving knowledge is acquired from both IL and model-based RL, where the agent can learn from human teachers as well as perform self-improvement by safely interacting with an offline world model. We validate our algorithm both in a high-fidelity driving simulation and on a real-world 1/20-scale RC-car with limited onboard computation. The evaluation results demonstrate that our method outperforms previous IL and RL methods in terms of sample efficiency and task performance. Demonstration videos are available at https://caipeide.github.io/autorace-dirl/

In many common-payoff games, achieving good performance requires players to develop protocols for communicating their private information implicitly -- i.e., using actions that have non-communicative effects on the environment. Multi-agent reinforcement learning practitioners typically approach this problem using independent learning methods in the hope that agents will learn implicit communication as a byproduct of expected return maximization. Unfortunately, independent learning methods are incapable of doing this in many settings. In this work, we isolate the implicit communication problem by identifying a class of partially observable common-payoff games, which we call implicit referential games, whose difficulty can be attributed to implicit communication. Next, we introduce a principled method based on minimum entropy coupling that leverages the structure of implicit referential games, yielding a new perspective on implicit communication. Lastly, we show that this method can discover performant implicit communication protocols in settings with very large spaces of messages.

Deep reinforcement learning has shown remarkable success in the past few years. Highly complex sequential decision making problems from game playing and robotics have been solved with deep model-free methods. Unfortunately, the sample complexity of model-free methods is often high. To reduce the number of environment samples, model-based reinforcement learning creates an explicit model of the environment dynamics. Achieving high model accuracy is a challenge in high-dimensional problems. In recent years, a diverse landscape of model-based methods has been introduced to improve model accuracy, using methods such as uncertainty modeling, model-predictive control, latent models, and end-to-end learning and planning. Some of these methods succeed in achieving high accuracy at low sample complexity, most do so either in a robotics or in a games context. In this paper, we survey these methods; we explain in detail how they work and what their strengths and weaknesses are. We conclude with a research agenda for future work to make the methods more robust and more widely applicable to other applications.