In this paper, we propose a Q-learning based decision-making framework to improve the safety and efficiency of Autonomous Vehicles when they encounter other maliciously behaving vehicles while passing through unsignalized intersections. In Autonomous Vehicles, conventional reward signals are set as regular rewards regarding feedback factors such as safety and efficiency. In this paper, safety gains are modulated by variable weighting parameters to ensure that safety can be emphasized more in emergency situations. The framework proposed in this paper introduces first-order theory of mind inferences on top of conventional rewards, using first-order beliefs as additional reward signals. The decision framework enables Autonomous Vehicles to make informed decisions when encountering vehicles with potentially malicious behaviors at unsignalized intersections, thereby improving the overall safety and efficiency of Autonomous Vehicle transportation systems. In order to verify the performance of the decision framework, this paper uses Prescan/Simulink co-simulations for simulation, and the results show that the performance of the decision framework can meet the set requirements.
The traditional approach to POMDPs is to convert them into fully observed MDPs by considering a belief state as an information state. However, a belief-state based approach requires perfect knowledge of the system dynamics and is therefore not applicable in the learning setting where the system model is unknown. Various approaches to circumvent this limitation have been proposed in the literature. We present a unified treatment of some of these approaches by viewing them as models where the agent maintains a local recursively updateable agent state and chooses actions based on the agent state. We highlight the different classes of agent-state based policies and the various approaches that have been proposed in the literature to find good policies within each class. These include the designer's approach to find optimal non-stationary agent-state based policies, policy search approaches to find a locally optimal stationary agent-state based policies, and the approximate information state to find approximately optimal stationary agent-state based policies. We then present how ideas from the approximate information state approach have been used to improve Q-learning and actor-critic algorithms for learning in POMDPs.
Q-learning is a powerful tool for network control and policy optimization in wireless networks, but it struggles with large state spaces. Recent advancements, like multi-environment mixed Q-learning (MEMQ), improves performance and reduces complexity by integrating multiple Q-learning algorithms across multiple related environments so-called digital cousins. However, MEMQ is designed for centralized single-agent networks and is not suitable for decentralized or multi-agent networks. To address this challenge, we propose a novel multi-agent MEMQ algorithm for partially decentralized wireless networks with multiple mobile transmitters (TXs) and base stations (BSs), where TXs do not have access to each other's states and actions. In uncoordinated states, TXs act independently to minimize their individual costs. In coordinated states, TXs use a Bayesian approach to estimate the joint state based on local observations and share limited information with leader TX to minimize joint cost. The cost of information sharing scales linearly with the number of TXs and is independent of the joint state-action space size. The proposed scheme is 50% faster than centralized MEMQ with only a 20% increase in average policy error (APE) and is 25% faster than several advanced decentralized Q-learning algorithms with 40% less APE. The convergence of the algorithm is also demonstrated.
Monte Carlo Tree Search (MCTS) is a powerful algorithm for solving complex decision-making problems. This paper presents an optimized MCTS implementation applied to the FrozenLake environment, a classic reinforcement learning task characterized by stochastic transitions. The optimization leverages cumulative reward and visit count tables along with the Upper Confidence Bound for Trees (UCT) formula, resulting in efficient learning in a slippery grid world. We benchmark our implementation against other decision-making algorithms, including MCTS with Policy and Q-Learning, and perform a detailed comparison of their performance. The results demonstrate that our optimized approach effectively maximizes rewards and success rates while minimizing convergence time, outperforming baseline methods, especially in environments with inherent randomness.
Video game playing is an extremely structured domain where algorithmic decision-making can be tested without adverse real-world consequences. While prevailing methods rely on image inputs to avoid the problem of hand-crafting state space representations, this approach systematically diverges from the way humans actually learn to play games. In this paper, we design object-based input representations that generalize well across a number of video games. Using these representations, we evaluate an agent's ability to learn games similar to an infant - with limited world experience, employing simple inductive biases derived from intuitive representations of physics from the real world. Using such biases, we construct an object category representation to be used by a Q-learning algorithm and assess how well it learns to play multiple games based on observed object affordances. Our results suggest that a human-like object interaction setup capably learns to play several video games, and demonstrates superior generalizability, particularly for unfamiliar objects. Further exploring such methods will allow machines to learn in a human-centric way, thus incorporating more human-like learning benefits.
Reinforcement learning (RL) has seen significant research and application results but often requires large amounts of training data. This paper proposes two data-efficient off-policy RL methods that use parametrized Q-learning. In these methods, the Q-function is chosen to be linear in the parameters and quadratic in selected basis functions in the state and control deviations from a base policy. A cost penalizing the $\ell_1$-norm of Bellman errors is minimized. We propose two methods: Linear Matrix Inequality Q-Learning (LMI-QL) and its iterative variant (LMI-QLi), which solve the resulting episodic optimization problem through convex optimization. LMI-QL relies on a convex relaxation that yields a semidefinite programming (SDP) problem with linear matrix inequalities (LMIs). LMI-QLi entails solving sequential iterations of an SDP problem. Both methods combine convex optimization with direct Q-function learning, significantly improving learning speed. A numerical case study demonstrates their advantages over existing parametrized Q-learning methods.
Proton pencil beam scanning (PBS) treatment planning for head and neck (H&N) cancers is a time-consuming and experience-demanding task where a large number of planning objectives are involved. Deep reinforcement learning (DRL) has recently been introduced to the planning processes of intensity-modulated radiation therapy and brachytherapy for prostate, lung, and cervical cancers. However, existing approaches are built upon the Q-learning framework and weighted linear combinations of clinical metrics, suffering from poor scalability and flexibility and only capable of adjusting a limited number of planning objectives in discrete action spaces. We propose an automatic treatment planning model using the proximal policy optimization (PPO) algorithm and a dose distribution-based reward function for proton PBS treatment planning of H&N cancers. Specifically, a set of empirical rules is used to create auxiliary planning structures from target volumes and organs-at-risk (OARs), along with their associated planning objectives. These planning objectives are fed into an in-house optimization engine to generate the spot monitor unit (MU) values. A decision-making policy network trained using PPO is developed to iteratively adjust the involved planning objective parameters in a continuous action space and refine the PBS treatment plans using a novel dose distribution-based reward function. Proton H&N treatment plans generated by the model show improved OAR sparing with equal or superior target coverage when compared with human-generated plans. Moreover, additional experiments on liver cancer demonstrate that the proposed method can be successfully generalized to other treatment sites. To the best of our knowledge, this is the first DRL-based automatic treatment planning model capable of achieving human-level performance for H&N cancers.
The Job Shop Scheduling Problem (JSSP) is a complex combinatorial optimization problem. While online Reinforcement Learning (RL) has shown promise by quickly finding acceptable solutions for JSSP, it faces key limitations: it requires extensive training interactions from scratch leading to sample inefficiency, cannot leverage existing high-quality solutions, and often yields suboptimal results compared to traditional methods like Constraint Programming (CP). We introduce Offline Reinforcement Learning for Learning to Dispatch (Offline-LD), which addresses these limitations by learning from previously generated solutions. Our approach is motivated by scenarios where historical scheduling data and expert solutions are available, although our current evaluation focuses on benchmark problems. Offline-LD adapts two CQL-based Q-learning methods (mQRDQN and discrete mSAC) for maskable action spaces, introduces a novel entropy bonus modification for discrete SAC, and exploits reward normalization through preprocessing. Our experiments demonstrate that Offline-LD outperforms online RL on both generated and benchmark instances. Notably, by introducing noise into the expert dataset, we achieve similar or better results than those obtained from the expert dataset, suggesting that a more diverse training set is preferable because it contains counterfactual information.
The ability of neural networks to perform robotic perception and control tasks such as depth and optical flow estimation, simultaneous localization and mapping (SLAM), and automatic control has led to their widespread adoption in recent years. Deep Reinforcement Learning has been used extensively in these settings, as it does not have the unsustainable training costs associated with supervised learning. However, DeepRL suffers from poor sample efficiency, i.e., it requires a large number of environmental interactions to converge to an acceptable solution. Modern RL algorithms such as Deep Q Learning and Soft Actor-Critic attempt to remedy this shortcoming but can not provide the explainability required in applications such as autonomous robotics. Humans intuitively understand the long-time-horizon sequential tasks common in robotics. Properly using such intuition can make RL policies more explainable while enhancing their sample efficiency. In this work, we propose SHIRE, a novel framework for encoding human intuition using Probabilistic Graphical Models (PGMs) and using it in the Deep RL training pipeline to enhance sample efficiency. Our framework achieves 25-78% sample efficiency gains across the environments we evaluate at negligible overhead cost. Additionally, by teaching RL agents the encoded elementary behavior, SHIRE enhances policy explainability. A real-world demonstration further highlights the efficacy of policies trained using our framework.
Kolmogorov-Arnold Networks (KAN) is an emerging neural network architecture in machine learning. It has greatly interested the research community about whether KAN can be a promising alternative of the commonly used Multi-Layer Perceptions (MLP). Experiments in various fields demonstrated that KAN-based machine learning can achieve comparable if not better performance than MLP-based methods, but with much smaller parameter scales and are more explainable. In this paper, we explore the incorporation of KAN into the actor and critic networks for offline reinforcement learning (RL). We evaluated the performance, parameter scales, and training efficiency of various KAN and MLP based conservative Q-learning (CQL) on the the classical D4RL benchmark for offline RL. Our study demonstrates that KAN can achieve performance close to the commonly used MLP with significantly fewer parameters. This provides us an option to choose the base networks according to the requirements of the offline RL tasks.
Although deep reinforcement learning (DRL) approaches in audio signal processing have seen substantial progress in recent years, audio-driven DRL for tasks such as navigation, gaze control and head-orientation control in the context of human-robot interaction have received little attention. Here, we propose an audio-driven DRL framework in which we utilise deep Q-learning to develop an autonomous agent that orients towards a talker in the acoustic environment based on stereo speech recordings. Our results show that the agent learned to perform the task at a near perfect level when trained on speech segments in anechoic environments (that is, without reverberation). The presence of reverberation in naturalistic acoustic environments affected the agent's performance, although the agent still substantially outperformed a baseline, randomly acting agent. Finally, we quantified the degree of generalization of the proposed DRL approach across naturalistic acoustic environments. Our experiments revealed that policies learned by agents trained on medium or high reverb environments generalized to low reverb environments, but policies learned by agents trained on anechoic or low reverb environments did not generalize to medium or high reverb environments. Taken together, this study demonstrates the potential of audio-driven DRL for tasks such as head-orientation control and highlights the need for training strategies that enable robust generalization across environments for real-world audio-driven DRL applications.
Optical wireless communication (OWC) is envisioned as one of the main enabling technologies of 6G networks, complementing radio frequency (RF) systems to provide high data rates. One of the crucial issues in indoor OWC is service interruptions due to blockages that obstruct the line of sight (LoS) between users and their access points (APs). Recently, reflecting surfaces referred to as intelligent reflecting surfaces (IRSs) have been considered to provide improved connectivity in OWC systems by reflecting AP signals toward users. In this study, we investigate the integration of IRSs into an indoor OWC system to improve the sum rate of the users and to ensure service continuity. We formulate an optimization problem for sum rate maximization, where the allocation of both APs and mirror elements of IRSs to users is determined to enhance the aggregate data rate. Moreover, reinforcement learning (RL) algorithms, specifically Q-learning and SARSA algorithms, are proposed to provide real-time solutions with low complexity and without prior system knowledge. The results show that the RL algorithms achieve near-optimal solutions that are close to the solutions of mixed integer linear programming (MILP). The results also show that the proposed scheme achieves up to a 45% increase in data rate compared to a traditional scheme that optimizes only the allocation of APs while the mirror elements are assigned to users based on the distance.
We study the Whittle index learning algorithm for restless multi-armed bandits (RMAB). We first present Q-learning algorithm and its variants -- speedy Q-learning (SQL), generalized speedy Q-learning (GSQL) and phase Q-learning (PhaseQL). We also discuss exploration policies -- $\epsilon$-greedy and Upper confidence bound (UCB). We extend the study of Q-learning and its variants with UCB policy. We illustrate using numerical example that Q-learning with UCB exploration policy has faster convergence and PhaseQL with UCB have fastest convergence rate. We next extend the study of Q-learning variants for index learning to RMAB. The algorithm of index learning is two-timescale variant of stochastic approximation, on slower timescale we update index learning scheme and on faster timescale we update Q-learning assuming fixed index value. We study constant stepsizes two timescale stochastic approximation algorithm. We describe the performance of our algorithms using numerical example. It illustrate that index learning with Q learning with UCB has faster convergence that $\epsilon$ greedy. Further, PhaseQL (with UCB and $\epsilon$ greedy) has the best convergence than other Q-learning algorithms.
Q-learning is a widely used algorithm in reinforcement learning (RL), but its convergence can be slow, especially when the discount factor is close to one. Successive Over-Relaxation (SOR) Q-learning, which introduces a relaxation factor to speed up convergence, addresses this issue but has two major limitations: In the tabular setting, the relaxation parameter depends on transition probability, making it not entirely model-free, and it suffers from overestimation bias. To overcome these limitations, we propose a sample-based, model-free double SOR Q-learning algorithm. Theoretically and empirically, this algorithm is shown to be less biased than SOR Q-learning. Further, in the tabular setting, the convergence analysis under boundedness assumptions on iterates is discussed. The proposed algorithm is extended to large-scale problems using deep RL. Finally, the tabular version of the proposed algorithm is compared using roulette and grid world environments, while the deep RL version is tested on a maximization bias example and OpenAI Gym environments.
We study the Whittle index learning algorithm for restless multi-armed bandits. We consider index learning algorithm with Q-learning. We first present Q-learning algorithm with exploration policies -- epsilon-greedy, softmax, epsilon-softmax with constant stepsizes. We extend the study of Q-learning to index learning for single-armed restless bandit. The algorithm of index learning is two-timescale variant of stochastic approximation, on slower timescale we update index learning scheme and on faster timescale we update Q-learning assuming fixed index value. In Q-learning updates are in asynchronous manner. We study constant stepsizes two timescale stochastic approximation algorithm. We provide analysis of two-timescale stochastic approximation for index learning with constant stepsizes. Further, we present study on index learning with deep Q-network (DQN) learning and linear function approximation with state-aggregation method. We describe the performance of our algorithms using numerical examples. We have shown that index learning with Q learning, DQN and function approximations learns the Whittle index.
We propose a new reinforcement learning (RL) formulation for training continuous-time score-based diffusion models for generative AI to generate samples that maximize reward functions while keeping the generated distributions close to the unknown target data distributions. Different from most existing studies, our formulation does not involve any pretrained model for the unknown score functions of the noise-perturbed data distributions. We present an entropy-regularized continuous-time RL problem and show that the optimal stochastic policy has a Gaussian distribution with a known covariance matrix. Based on this result, we parameterize the mean of Gaussian policies and develop an actor-critic type (little) q-learning algorithm to solve the RL problem. A key ingredient in our algorithm design is to obtain noisy observations from the unknown score function via a ratio estimator. Numerically, we show the effectiveness of our approach by comparing its performance with two state-of-the-art RL methods that fine-tune pretrained models. Finally, we discuss extensions of our RL formulation to probability flow ODE implementation of diffusion models and to conditional diffusion models.
This paper studies asynchronous stochastic approximation (SA) algorithms and their theoretical application to reinforcement learning in semi-Markov decision processes (SMDPs) with an average-reward criterion. We first extend Borkar and Meyn's stability proof method to accommodate more general noise conditions, yielding broader convergence guarantees for asynchronous SA. To sharpen the convergence analysis, we further examine shadowing properties in the asynchronous setting, building on a dynamical-systems approach of Hirsch and Bena\"{i}m. Leveraging these SA results, we establish the convergence of an asynchronous SA analogue of Schweitzer's classical relative value iteration algorithm, RVI Q-learning, for finite-space, weakly communicating SMDPs. Moreover, to make full use of these SA results in this application, we introduce new monotonicity conditions for estimating the optimal reward rate in RVI Q-learning. These conditions substantially expand the previously considered algorithmic framework, and we address them with novel arguments in the stability and convergence analysis of RVI Q-learning.
Large-scale multi-agent systems are often deployed across wide geographic areas, where agents interact with heterogeneous environments. There is an emerging interest in understanding the role of heterogeneity in the performance of the federated versions of classic reinforcement learning algorithms. In this paper, we study synchronous federated Q-learning, which aims to learn an optimal Q-function by having $K$ agents average their local Q-estimates per $E$ iterations. We observe an interesting phenomenon on the convergence speeds in terms of $K$ and $E$. Similar to the homogeneous environment settings, there is a linear speed-up concerning $K$ in reducing the errors that arise from sampling randomness. Yet, in sharp contrast to the homogeneous settings, $E>1$ leads to significant performance degradation. Specifically, we provide a fine-grained characterization of the error evolution in the presence of environmental heterogeneity, which decay to zero as the number of iterations $T$ increases. The slow convergence of having $E>1$ turns out to be fundamental rather than an artifact of our analysis. We prove that, for a wide range of stepsizes, the $\ell_{\infty}$ norm of the error cannot decay faster than $\Theta (E/T)$. In addition, our experiments demonstrate that the convergence exhibits an interesting two-phase phenomenon. For any given stepsize, there is a sharp phase-transition of the convergence: the error decays rapidly in the beginning yet later bounces up and stabilizes. Provided that the phase-transition time can be estimated, choosing different stepsizes for the two phases leads to faster overall convergence.
Recently, there has been a surge of interest in analyzing the non-asymptotic behavior of model-free reinforcement learning algorithms. However, the performance of such algorithms in non-ideal environments, such as in the presence of corrupted rewards, is poorly understood. Motivated by this gap, we investigate the robustness of the celebrated Q-learning algorithm to a strong-contamination attack model, where an adversary can arbitrarily perturb a small fraction of the observed rewards. We start by proving that such an attack can cause the vanilla Q-learning algorithm to incur arbitrarily large errors. We then develop a novel robust synchronous Q-learning algorithm that uses historical reward data to construct robust empirical Bellman operators at each time step. Finally, we prove a finite-time convergence rate for our algorithm that matches known state-of-the-art bounds (in the absence of attacks) up to a small inevitable $O(\varepsilon)$ error term that scales with the adversarial corruption fraction $\varepsilon$. Notably, our results continue to hold even when the true reward distributions have infinite support, provided they admit bounded second moments.
Robots find extensive applications in industry. In recent years, the influence of robots has also increased rapidly in domestic scenarios. The Q-learning algorithm aims to maximise the reward for reaching the goal. This paper proposes a modified version of the Q-learning algorithm, known as Q-learning with scaled distance metric (Q-SD). This algorithm enhances task learning and makes task completion more meaningful. A robotic manipulator (agent) applies the Q-SD algorithm to the task of table cleaning. Using Q-SD, the agent acquires the sequence of steps necessary to accomplish the task while minimising the manipulator's movement distance. We partition the table into grids of different dimensions. The first has a grid count of 3 times 3, and the second has a grid count of 4 times 4. Using the Q-SD algorithm, the maximum success obtained in these two environments was 86% and 59% respectively. Moreover, Compared to the conventional Q-learning algorithm, the drop in average distance moved by the agent in these two environments using the Q-SD algorithm was 8.61% and 6.7% respectively.