The Metaverse is emerging as maturing technologies are empowering the different facets. Virtual Reality (VR) technologies serve as the backbone of the virtual universe within the Metaverse to offer a highly immersive user experience. As mobility is emphasized in the Metaverse context, VR devices reduce their weights at the sacrifice of local computation abilities. In this paper, for a system consisting of a Metaverse server and multiple VR users, we consider two cases of (i) the server generating frames and transmitting them to users, and (ii) users generating frames locally and thus consuming device energy. Moreover, in our multi-user VR scenario for the Metaverse, users have different characteristics and demands for Frames Per Second (FPS). Then the channel access arrangement (including the decisions on frame generation location), and transmission powers for the downlink communications from the server to the users are jointly optimized to improve the utilities of users. This joint optimization is addressed by deep reinforcement learning (DRL) with heterogeneous actions. Our proposed user-centric DRL algorithm is called User-centric Critic with Heterogenous Actors (UCHA). Extensive experiments demonstrate that our UCHA algorithm leads to remarkable results under various requirements and constraints.

The ability to pick up on language signals in an ongoing interaction is crucial for future machine learning models to collaborate and interact with humans naturally. In this paper, we present an initial study that evaluates intraepisodic feedback given in a collaborative setting. We use a referential language game as a controllable example of a task-oriented collaborative joint activity. A teacher utters a referring expression generated by a well-known symbolic algorithm (the "Incremental Algorithm") as an initial instruction and then monitors the follower's actions to possibly intervene with intra-episodic feedback (which does not explicitly have to be requested). We frame this task as a reinforcement learning problem with sparse rewards and learn a follower policy for a heuristic teacher. Our results show that intra-episodic feedback allows the follower to generalize on aspects of scene complexity and Figure 1: An exemplary interaction between a teacher performs better than providing only the initial and a follower that controls the gripper (the grey statement.

Quantum machine learning implemented by variational quantum circuits (VQCs) is considered a promising concept for the noisy intermediate-scale quantum computing era. Focusing on applications in quantum reinforcement learning, we propose a specific action decoding procedure for a quantum policy gradient approach. We introduce a novel quality measure that enables us to optimize the classical post-processing required for action selection, inspired by local and global quantum measurements. The resulting algorithm demonstrates a significant performance improvement in several benchmark environments. With this technique, we successfully execute a full training routine on a 5-qubit hardware device. Our method introduces only negligible classical overhead and has the potential to improve VQC-based algorithms beyond the field of quantum reinforcement learning.

Importance sampling (IS) is a popular technique in off-policy evaluation, which re-weights the return of trajectories in the replay buffer to boost sample efficiency. However, training with IS can be unstable and previous attempts to address this issue mainly focus on analyzing the variance of IS. In this paper, we reveal that the instability is also related to a new notion of Reuse Bias of IS -- the bias in off-policy evaluation caused by the reuse of the replay buffer for evaluation and optimization. We theoretically show that the off-policy evaluation and optimization of the current policy with the data from the replay buffer result in an overestimation of the objective, which may cause an erroneous gradient update and degenerate the performance. We further provide a high-probability upper bound of the Reuse Bias, and show that controlling one term of the upper bound can control the Reuse Bias by introducing the concept of stability for off-policy algorithms. Based on these analyses, we finally present a novel Bias-Regularized Importance Sampling (BIRIS) framework along with practical algorithms, which can alleviate the negative impact of the Reuse Bias. Experimental results show that our BIRIS-based methods can significantly improve the sample efficiency on a series of continuous control tasks in MuJoCo.

In recent years, the development of Artificial Intelligence (AI) has shown tremendous potential in diverse areas. Among them, reinforcement learning (RL) has proven to be an effective solution for learning intelligent control strategies. As an inevitable trend for mitigating climate change, hybrid electric vehicles (HEVs) rely on efficient energy management strategies (EMS) to minimize energy consumption. Many researchers have employed RL to learn optimal EMS for specific vehicle models. However, most of these models tend to be complex and proprietary, making them unsuitable for broad applicability. This paper presents a novel framework, in which we implement and integrate RL-based EMS with the open-source vehicle simulation tool called FASTSim. The learned RL-based EMSs are evaluated on various vehicle models using different test drive cycles and prove to be effective in improving energy efficiency.

Standard model-based reinforcement learning (MBRL) approaches fit a transition model of the environment to all past experience, but this wastes model capacity on data that is irrelevant for policy improvement. We instead propose a new "transition occupancy matching" (TOM) objective for MBRL model learning: a model is good to the extent that the current policy experiences the same distribution of transitions inside the model as in the real environment. We derive TOM directly from a novel lower bound on the standard reinforcement learning objective. To optimize TOM, we show how to reduce it to a form of importance weighted maximum-likelihood estimation, where the automatically computed importance weights identify policy-relevant past experiences from a replay buffer, enabling stable optimization. TOM thus offers a plug-and-play model learning sub-routine that is compatible with any backbone MBRL algorithm. On various Mujoco continuous robotic control tasks, we show that TOM successfully focuses model learning on policy-relevant experience and drives policies faster to higher task rewards than alternative model learning approaches. Code can be found on our project website: penn-pal-lab.github.io/TOM/

We study variance-dependent regret bounds for Markov decision processes (MDPs). Algorithms with variance-dependent regret guarantees can automatically exploit environments with low variance (e.g., enjoying constant regret on deterministic MDPs). The existing algorithms are either variance-independent or suboptimal. We first propose two new environment norms to characterize the fine-grained variance properties of the environment. For model-based methods, we design a variant of the MVP algorithm (Zhang et al., 2021a). We apply new analysis techniques to demonstrate that this algorithm enjoys variance-dependent bounds with respect to the norms we propose. In particular, this bound is simultaneously minimax optimal for both stochastic and deterministic MDPs, the first result of its kind. We further initiate the study on model-free algorithms with variance-dependent regret bounds by designing a reference-function-based algorithm with a novel capped-doubling reference update schedule. Lastly, we also provide lower bounds to complement our upper bounds.

We study regret minimization for reinforcement learning (RL) in Latent Markov Decision Processes (LMDPs) with context in hindsight. We design a novel model-based algorithmic framework which can be instantiated with both a model-optimistic and a value-optimistic solver. We prove an $\tilde{O}(\sqrt{\mathsf{Var}^\star M \Gamma S A K})$ regret bound where $\tilde{O}$ hides logarithm factors, $M$ is the number of contexts, $S$ is the number of states, $A$ is the number of actions, $K$ is the number of episodes, $\Gamma \le S$ is the maximum transition degree of any state-action pair, and $\mathsf{Var}^\star$ is a variance quantity describing the determinism of the LMDP. The regret bound only scales logarithmically with the planning horizon, thus yielding the first (nearly) horizon-free regret bound for LMDP. This is also the first problem-dependent regret bound for LMDP. Key in our proof is an analysis of the total variance of alpha vectors (a generalization of value functions), which is handled with a truncation method. We complement our positive result with a novel $\Omega(\sqrt{\mathsf{Var}^\star M S A K})$ regret lower bound with $\Gamma = 2$, which shows our upper bound minimax optimal when $\Gamma$ is a constant for the class of variance-bounded LMDPs. Our lower bound relies on new constructions of hard instances and an argument inspired by the symmetrization technique from theoretical computer science, both of which are technically different from existing lower bound proof for MDPs, and thus can be of independent interest.

In this work, we present an approach to supervisory reinforcement learning control for unmanned aerial vehicles (UAVs). UAVs are dynamic systems where control decisions in response to disturbances in the environment have to be made in the order of milliseconds. We formulate a supervisory control architecture that interleaves with extant embedded control and demonstrates robustness to environmental disturbances in the form of adverse wind conditions. We run case studies with a Tarot T-18 Octorotor to demonstrate the effectiveness of our approach and compare it against a classic cascade control architecture used in most vehicles. While the results show the performance difference is marginal for nominal operations, substantial performance improvement is obtained with the supervisory RL approach under unseen wind conditions.

In this paper, we leverage ideas from model-based control to address the sample efficiency problem of reinforcement learning (RL) algorithms. Accelerating learning is an active field of RL highly relevant in the context of time-varying systems. Traditional transfer learning methods propose to use prior knowledge of the system behavior to devise a gradual or immediate data-driven transformation of the control policy obtained through RL. Such transformation is usually computed by estimating the performance of previous control policies based on measurements recently collected from the system. However, such retrospective measures have debatable utility with no guarantees of positive transfer in most cases. Instead, we propose a model-based transformation, such that when actions from a control policy are applied to the target system, a positive transfer is achieved. The transformation can be used as an initialization for the reinforcement learning process to converge to a new optimum. We validate the performance of our approach through four benchmark examples. We demonstrate that our approach is more sample-efficient than fine-tuning with reinforcement learning alone and achieves comparable performance to linear-quadratic-regulators and model-predictive control when an accurate linear model is known in the three cases. If an accurate model is not known, we empirically show that the proposed approach still guarantees positive transfer with jump-start improvement.