Offline reinforcement learning (RL) enables the agent to effectively learn from logged data, which significantly extends the applicability of RL algorithms in real-world scenarios where exploration can be expensive or unsafe. Previous works have shown that extracting primitive skills from the recurring and temporally extended structures in the logged data yields better learning. However, these methods suffer greatly when the primitives have limited representation ability to recover the original policy space, especially in offline settings. In this paper, we give a quantitative characterization of the performance of offline hierarchical learning and highlight the importance of learning lossless primitives. To this end, we propose to use a \emph{flow}-based structure as the representation for low-level policies. This allows us to represent the behaviors in the dataset faithfully while keeping the expression ability to recover the whole policy space. We show that such lossless primitives can drastically improve the performance of hierarchical policies. The experimental results and extensive ablation studies on the standard D4RL benchmark show that our method has a good representation ability for policies and achieves superior performance in most tasks.

In constrained reinforcement learning (C-RL), an agent seeks to learn from the environment a policy that maximizes the expected cumulative reward while satisfying minimum requirements in secondary cumulative reward constraints. Several algorithms rooted in sampled-based primal-dual methods have been recently proposed to solve this problem in policy space. However, such methods are based on stochastic gradient descent ascent algorithms whose trajectories are connected to the optimal policy only after a mixing output stage that depends on the algorithm's history. As a result, there is a mismatch between the behavioral policy and the optimal one. In this work, we propose a novel algorithm for constrained RL that does not suffer from these limitations. Leveraging recent results on regularized saddle-flow dynamics, we develop a novel stochastic gradient descent-ascent algorithm whose trajectories converge to the optimal policy almost surely.

Deep Reinforcement Learning (DRL) has the potential to be used for synthesizing feedback controllers (agents) for various complex systems with unknown dynamics. These systems are expected to satisfy diverse safety and liveness properties best captured using temporal logic. In RL, the reward function plays a crucial role in specifying the desired behaviour of these agents. However, the problem of designing the reward function for an RL agent to satisfy complex temporal logic specifications has received limited attention in the literature. To address this, we provide a systematic way of generating rewards in real-time by using the quantitative semantics of Signal Temporal Logic (STL), a widely used temporal logic to specify the behaviour of cyber-physical systems. We propose a new quantitative semantics for STL having several desirable properties, making it suitable for reward generation. We evaluate our STL-based reinforcement learning mechanism on several complex continuous control benchmarks and compare our STL semantics with those available in the literature in terms of their efficacy in synthesizing the controller agent. Experimental results establish our new semantics to be the most suitable for synthesizing feedback controllers for complex continuous dynamical systems through reinforcement learning.

To make machine learning (ML) sustainable and apt to run on the diverse devices where relevant data is, it is essential to compress ML models as needed, while still meeting the required learning quality and time performance. However, how much and when an ML model should be compressed, and {\em where} its training should be executed, are hard decisions to make, as they depend on the model itself, the resources of the available nodes, and the data such nodes own. Existing studies focus on each of those aspects individually, however, they do not account for how such decisions can be made jointly and adapted to one another. In this work, we model the network system focusing on the training of DNNs, formalize the above multi-dimensional problem, and, given its NP-hardness, formulate an approximate dynamic programming problem that we solve through the PACT algorithmic framework. Importantly, PACT leverages a time-expanded graph representing the learning process, and a data-driven and theoretical approach for the prediction of the loss evolution to be expected as a consequence of training decisions. We prove that PACT's solutions can get as close to the optimum as desired, at the cost of an increased time complexity, and that, in any case, such complexity is polynomial. Numerical results also show that, even under the most disadvantageous settings, PACT outperforms state-of-the-art alternatives and closely matches the optimal energy cost.

We consider the problem of autonomous channel access (AutoCA), where a group of terminals tries to discover a communication strategy with an access point (AP) via a common wireless channel in a distributed fashion. Due to the irregular topology and the limited communication range of terminals, a practical challenge for AutoCA is the hidden terminal problem, which is notorious in wireless networks for deteriorating the throughput and delay performances. To meet the challenge, this paper presents a new multi-agent deep reinforcement learning paradigm, dubbed MADRL-HT, tailored for AutoCA in the presence of hidden terminals. MADRL-HT exploits topological insights and transforms the observation space of each terminal into a scalable form independent of the number of terminals. To compensate for the partial observability, we put forth a look-back mechanism such that the terminals can infer behaviors of their hidden terminals from the carrier sensed channel states as well as feedback from the AP. A window-based global reward function is proposed, whereby the terminals are instructed to maximize the system throughput while balancing the terminals' transmission opportunities over the course of learning. Extensive numerical experiments verified the superior performance of our solution benchmarked against the legacy carrier-sense multiple access with collision avoidance (CSMA/CA) protocol.

This paper studies a distributed policy gradient in collaborative multi-agent reinforcement learning (MARL), where agents over a communication network aim to find the optimal policy to maximize the average of all agents' local returns. Due to the non-concave performance function of policy gradient, the existing distributed stochastic optimization methods for convex problems cannot be directly used for policy gradient in MARL. This paper proposes a distributed policy gradient with variance reduction and gradient tracking to address the high variances of policy gradient, and utilizes importance weight to solve the {distribution shift} problem in the sampling process. We then provide an upper bound on the mean-squared stationary gap, which depends on the number of iterations, the mini-batch size, the epoch size, the problem parameters, and the network topology. We further establish the sample and communication complexity to obtain an $\epsilon$-approximate stationary point. Numerical experiments are performed to validate the effectiveness of the proposed algorithm.

The popular Proximal Policy Optimization (PPO) algorithm approximates the solution in a clipped policy space. Does there exist better policies outside of this space? By using a novel surrogate objective that employs the sigmoid function (which provides an interesting way of exploration), we found that the answer is ``YES'', and the better policies are in fact located very far from the clipped space. We show that PPO is insufficient in ``off-policyness'', according to an off-policy metric called DEON. Our algorithm explores in a much larger policy space than PPO, and it maximizes the Conservative Policy Iteration (CPI) objective better than PPO during training. To the best of our knowledge, all current PPO methods have the clipping operation and optimize in the clipped policy space. Our method is the first of this kind, which advances the understanding of CPI optimization and policy gradient methods. Code is available at https://github.com/raincchio/P3O.

DeepNash uses a novel approach, based on game theory and model-free deep reinforcement learning.

Achieving fast cooling of motional modes is a prerequisite for leveraging such bosonic quanta for high-speed quantum information processing. In this Letter, we address the aspect of reducing the time limit for cooling, below that constrained by the conventional sideband cooling techniques, and propose a scheme to apply deep reinforcement learning (DRL) to achieve this. In particular, we have numerically demonstrated how the scheme can be used effectively to accelerate the dynamic motional cooling of a macroscopic magnonic sphere, and how it can be uniformly extended to more complex systems, for example, a tripartite opto-magno-mechanical system, to obtain cooling of the motional mode below the time bound of coherent cooling. While conventional sideband cooling methods do not work beyond the well-known rotating wave approximation (RWA) regimes, our proposed DRL scheme can be applied uniformly to regimes operating within and beyond the RWA, and thus, this offers a new and complete toolkit for rapid control and generation of macroscopic quantum states for application in quantum technologies. Published by the American Physical Society under the terms of the Creative Commons Attribution 4.0 International license.

Safety constraints and optimality are important, but sometimes conflicting criteria for controllers. Although these criteria are often solved separately with different tools to maintain formal guarantees, it is also common practice in reinforcement learning to simply modify reward functions by penalizing failures, with the penalty treated as a mere heuristic. We rigorously examine the relationship of both safety and optimality to penalties, and formalize sufficient conditions for safe value functions (SVFs): value functions that are both optimal for a given task, and enforce safety constraints. We reveal this structure by examining when rewards preserve viability under optimal control, and show that there always exists a finite penalty that induces a safe value function. This penalty is not unique, but upper-unbounded: larger penalties do not harm optimality. Although it is often not possible to compute the minimum required penalty, we reveal clear structure of how the penalty, rewards, discount factor, and dynamics interact. This insight suggests practical, theory-guided heuristics to design reward functions for control problems where safety is important.