Recent years have seen remarkable successes of reinforcement learning (RL) in a variety of applications, such as board games (Silver et al., 2017), autonomous driving (Shalev-Shwartz et al., 2016), city navigation (Mirowski et al., 2018), and fusion plasma control (Degrave et al., 2022). A common feature of these applications is that there are multiple decision-makers interacting with each other in an unknown environment. While empirical successes have shown the potential of multi-agent reinforcement learning (MARL) (Busoniu et al., 2008; Zhang et al., 2021a), the training of MARL agents largely relies on heuristics and parameter-tuning, and is not always reliable. In particular, many practical MARL algorithms are heuristically extended from their single-agent counterparts and lack theoretical guarantees. A growing literature seeks to provide theoretical insights to substantiate the empirical success of MARL and inform the design of efficient, and provably convergent algorithms.

Multi-agent pursuit-evasion tasks involving intelligent targets are notoriously challenging coordination problems. In this paper, we investigate new ways to learn such coordinated behaviors of unmanned aerial vehicles (UAVs) aimed at keeping track of multiple evasive targets. Within a Multi-Agent Reinforcement Learning (MARL) framework, we specifically propose a variant of the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) method. Our approach addresses multi-target pursuit-evasion scenarios within non-stationary and unknown environments with random obstacles. In addition, given the critical role played by collective exploration in terms of detecting possible targets, we implement heterogeneous roles for the pursuers for enhanced exploratory actions balanced by exploitation (i.e. tracking) of previously identified targets. Our proposed role-based MADDPG algorithm is not only able to track multiple targets, but also is able to explore for possible targets by means of the proposed Voronoi-based rewarding policy. We implemented, tested and validated our approach in a simulation environment prior to deploying a real-world multi-robot system comprising of Crazyflie drones. Our results demonstrate that a multi-agent pursuit team has the ability to learn highly efficient coordinated control policies in terms of target tracking and exploration even when confronted with multiple fast evasive targets in complex environments.

Linear temporal logic (LTL) offers a simplified way of specifying tasks for policy optimization that may otherwise be difficult to describe with scalar reward functions. However, the standard RL framework can be too myopic to find maximally LTL satisfying policies. This paper makes two contributions. First, we develop a new value-function based proxy, using a technique we call eventual discounting, under which one can find policies that satisfy the LTL specification with highest achievable probability. Second, we develop a new experience replay method for generating off-policy data from on-policy rollouts via counterfactual reasoning on different ways of satisfying the LTL specification. Our experiments, conducted in both discrete and continuous state-action spaces, confirm the effectiveness of our counterfactual experience replay approach.

Offline reinforcement learning (RL) addresses the problem of learning a performant policy from a fixed batch of data collected by following some behavior policy. Model-based approaches are particularly appealing in the offline setting since they can extract more learning signals from the logged dataset by learning a model of the environment. However, the performance of existing model-based approaches falls short of model-free counterparts, due to the compounding of estimation errors in the learned model. Driven by this observation, we argue that it is critical for a model-based method to understand when to trust the model and when to rely on model-free estimates, and how to act conservatively w.r.t. both. To this end, we derive an elegant and simple methodology called conservative Bayesian model-based value expansion for offline policy optimization (CBOP), that trades off model-free and model-based estimates during the policy evaluation step according to their epistemic uncertainties, and facilitates conservatism by taking a lower bound on the Bayesian posterior value estimate. On the standard D4RL continuous control tasks, we find that our method significantly outperforms previous model-based approaches: e.g., MOPO by $116.4$%, MOReL by $23.2$% and COMBO by $23.7$%. Further, CBOP achieves state-of-the-art performance on $11$ out of $18$ benchmark datasets while doing on par on the remaining datasets.

Imitation learning aims to mimic the behavior of experts without explicit reward signals. Passive imitation learning methods which use static expert datasets typically suffer from compounding error, low sample efficiency, and high hyper-parameter sensitivity. In contrast, active imitation learning methods solicit expert interventions to address the limitations. However, recent active imitation learning methods are designed based on human intuitions or empirical experience without theoretical guarantee. In this paper, we propose a novel active imitation learning framework based on a teacher-student interaction model, in which the teacher's goal is to identify the best teaching behavior and actively affect the student's learning process. By solving the optimization objective of this framework, we propose a practical implementation, naming it AdapMen. Theoretical analysis shows that AdapMen can improve the error bound and avoid compounding error under mild conditions. Experiments on the MetaDrive benchmark and Atari 2600 games validate our theoretical analysis and show that our method achieves near-expert performance with much less expert involvement and total sampling steps than previous methods. The code is available at https://github.com/liuxhym/AdapMen.

The goal of Next-Generation Networks is to improve upon the current networking paradigm, especially in providing higher data rates, near-real-time latencies, and near-perfect quality of service. However, existing radio access network (RAN) architectures lack sufficient flexibility and intelligence to meet those demands. Open RAN (O-RAN) is a promising paradigm for building a virtualized and intelligent RAN architecture. This paper presents a Machine Learning (ML)-based Traffic Steering (TS) scheme to predict network congestion and then proactively steer O-RAN traffic to avoid it and reduce the expected queuing delay. To achieve this, we propose an optimized setup focusing on safeguarding both latency and reliability to serve URLLC applications. The proposed solution consists of a two-tiered ML strategy based on Naive Bayes Classifier and deep Q-learning. Our solution is evaluated against traditional reactive TS approaches that are offered as xApps in O-RAN and shows an average of 15.81 percent decrease in queuing delay across all deployed SFCs.

Deep reinforcement learning was instigated with the presence of trust region methods, being scalable and efficient. However, the pessimism of such algorithms, among which it forces to constrain in a trust region by all means, has been proven to suppress the exploration and harm the performance. Exploratory algorithm such as SAC, while utilizes the entropy to encourage exploration, implicitly optimizing another objective yet. We first observed this inconsistency, and therefore put forward an analogous augmentation technique, which combines well with the on-policy algorithms, when a value critic is involved. Surprisingly, the proposed method consistently satisfies the soft policy improvement theorem, while being more extensible. As the analysis advises, it is crucial to control the temperature coefficient to balance the exploration and exploitation. Empirical tests on MuJoCo benchmark tasks show that the agent is heartened towards higher reward regions, and enjoys a finer performance. Furthermore, we verify the exploration bonus of our method on a set of custom environments.

Using generative deep learning models and reinforcement learning together can effectively generate new molecules with desired properties. By employing a multi-objective scoring function, thousands of high-scoring molecules can be generated, making this approach useful for drug discovery and material science. However, the application of these methods can be hindered by computationally expensive or time-consuming scoring procedures, particularly when a large number of function calls are required as feedback in the reinforcement learning optimization. Here, we propose the use of double-loop reinforcement learning with simplified molecular line entry system (SMILES) augmentation to improve the efficiency and speed of the optimization. By adding an inner loop that augments the generated SMILES strings to non-canonical SMILES for use in additional reinforcement learning rounds, we can both reuse the scoring calculations on the molecular level, thereby speeding up the learning process, as well as offer additional protection against mode collapse. We find that employing between 5 and 10 augmentation repetitions is optimal for the scoring functions tested and is further associated with an increased diversity in the generated compounds, improved reproducibility of the sampling runs and the generation of molecules of higher similarity to known ligands.

The deployment of Reinforcement Learning to robotics applications faces the difficulty of reward engineering. Therefore, approaches have focused on creating reward functions by Learning from Observations (LfO) which is the task of learning policies from expert trajectories that only contain state sequences. We propose new methods for LfO for the important class of continuous control problems of learning to stabilize, by introducing intermediate proxy models acting as reward functions between the expert and the agent policy based on Lyapunov stability theory. Our LfO training process consists of two steps. The first step attempts to learn a Lyapunov-like landscape proxy model from expert state sequences without access to any kinematics model, and the second step uses the learned landscape model to guide in training the learner's policy. We formulate novel learning objectives for the two steps that are important for overall training success. We evaluate our methods in real automobile robot environments and other simulated stabilization control problems in model-free settings, like Quadrotor control and maintaining upright positions of Hopper in MuJoCo. We compare with state-of-the-art approaches and show the proposed methods can learn efficiently with less expert observations.

While robots have become increasingly advanced over the past few years, most of them are still unable to reliably navigate very crowded spaces, such as public areas or roads in urban environments. To be implemented on a large-scale and in the smart cities of the future, however, robots will need to be able to navigate these environments both reliably and safely, without colliding with humans or nearby objects. Researchers at the University of Zaragoza and the Aragon Institute of Engineering Research in Spain have recently proposed a new machine learning–based approach that could improve robot navigation in both indoor and outdoor crowded environments. This approach, introduced in a paper pre-published on the arXiv server, entails the use of intrinsic rewards, which are essentially "rewards" that an AI agent receives when performing behaviors that are not strictly related to the task it is trying to complete. "Autonomous robot navigation is an open unsolved problem, especially in unstructured and dynamic environments, where a robot has to avoid collisions with dynamic obstacles and reach the goal," Diego Martinez Baselga, one of the researchers who carried out the study, told Tech Xplore.