Traffic light control is important for reducing congestion in urban mobility systems. This paper proposes a real-time traffic light control method using deep Q learning. Our approach incorporates a reward function considering queue lengths, delays, travel time, and throughput. The model dynamically decides phase changes based on current traffic conditions. The training of the deep Q network involves an offline stage from pre-generated data with fixed schedules and an online stage using real-time traffic data. A deep Q network structure with a "phase gate" component is used to simplify the model's learning task under different phases. A "memory palace" mechanism is used to address sample imbalance during the training process. We validate our approach using both synthetic and real-world traffic flow data on a road intersecting in Hangzhou, China. Results demonstrate significant performance improvements of the proposed method in reducing vehicle waiting time (57.1% to 100%), queue lengths (40.9% to 100%), and total travel time (16.8% to 68.0%) compared to traditional fixed signal plans.
Cooperative Multi-agent Reinforcement Learning (MARL) algorithms with Zero-Shot Coordination (ZSC) have gained significant attention in recent years. ZSC refers to the ability of agents to coordinate zero-shot (without additional interaction experience) with independently trained agents. While ZSC is crucial for cooperative MARL agents, it might not be possible for complex tasks and changing environments. Agents also need to adapt and improve their performance with minimal interaction with other agents. In this work, we show empirically that state-of-the-art ZSC algorithms have poor performance when paired with agents trained with different learning methods, and they require millions of interaction samples to adapt to these new partners. To investigate this issue, we formally defined a framework based on a popular cooperative multi-agent game called Hanabi to evaluate the adaptability of MARL methods. In particular, we created a diverse set of pre-trained agents and defined a new metric called adaptation regret that measures the agent's ability to efficiently adapt and improve its coordination performance when paired with some held-out pool of partners on top of its ZSC performance. After evaluating several SOTA algorithms using our framework, our experiments reveal that naive Independent Q-Learning (IQL) agents in most cases adapt as quickly as the SOTA ZSC algorithm Off-Belief Learning (OBL). This finding raises an interesting research question: How to design MARL algorithms with high ZSC performance and capability of fast adaptation to unseen partners. As a first step, we studied the role of different hyper-parameters and design choices on the adaptability of current MARL algorithms. Our experiments show that two categories of hyper-parameters controlling the training data diversity and optimization process have a significant impact on the adaptability of Hanabi agents.
Blackjack or "21" is a popular card-based game of chance and skill. The objective of the game is to win by obtaining a hand total higher than the dealer's without exceeding 21. The ideal blackjack strategy will maximize financial return in the long run while avoiding gambler's ruin. The stochastic environment and inherent reward structure of blackjack presents an appealing problem to better understand reinforcement learning agents in the presence of environment variations. Here we consider a q-learning solution for optimal play and investigate the rate of learning convergence of the algorithm as a function of deck size. A blackjack simulator allowing for universal blackjack rules is also implemented to demonstrate the extent to which a card counter perfectly using the basic strategy and hi-lo system can bring the house to bankruptcy and how environment variations impact this outcome. The novelty of our work is to place this conceptual understanding of the impact of deck size in the context of learning agent convergence.
In real-world multi-agent reinforcement learning (MARL) applications, agents may not have perfect state information (e.g., due to inaccurate measurement or malicious attacks), which challenges the robustness of agents' policies. Though robustness is getting important in MARL deployment, little prior work has studied state uncertainties in MARL, neither in problem formulation nor algorithm design. Motivated by this robustness issue and the lack of corresponding studies, we study the problem of MARL with state uncertainty in this work. We provide the first attempt to the theoretical and empirical analysis of this challenging problem. We first model the problem as a Markov Game with state perturbation adversaries (MG-SPA) by introducing a set of state perturbation adversaries into a Markov Game. We then introduce robust equilibrium (RE) as the solution concept of an MG-SPA. We conduct a fundamental analysis regarding MG-SPA such as giving conditions under which such a robust equilibrium exists. Then we propose a robust multi-agent Q-learning (RMAQ) algorithm to find such an equilibrium, with convergence guarantees. To handle high-dimensional state-action space, we design a robust multi-agent actor-critic (RMAAC) algorithm based on an analytical expression of the policy gradient derived in the paper. Our experiments show that the proposed RMAQ algorithm converges to the optimal value function; our RMAAC algorithm outperforms several MARL and robust MARL methods in multiple multi-agent environments when state uncertainty is present. The source code is public on \url{https://github.com/sihongho/robust_marl_with_state_uncertainty}.
We leverage the fast physics simulator, MuJoCo to run tasks in a continuous control environment and reveal details like the observation space, action space, rewards, etc. for each task. We benchmark value-based methods for continuous control by comparing Q-learning and SARSA through a discretization approach, and using them as baselines, progressively moving into one of the state-of-the-art deep policy gradient method DDPG. Over a large number of episodes, Qlearning outscored SARSA, but DDPG outperformed both in a small number of episodes. Lastly, we also fine-tuned the model hyper-parameters expecting to squeeze more performance but using lesser time and resources. We anticipated that the new design for DDPG would vastly improve performance, yet after only a few episodes, we were able to achieve decent average rewards. We expect to improve the performance provided adequate time and computational resources.
Gradient-based learning in multi-agent systems is difficult because the gradient derives from a first-order model which does not account for the interaction between agents' learning processes. LOLA (arXiv:1709.04326) accounts for this by differentiating through one step of optimization. We propose to judge joint policies by their long-term prospects as measured by the meta-value, a discounted sum over the returns of future optimization iterates. We apply a form of Q-learning to the meta-game of optimization, in a way that avoids the need to explicitly represent the continuous action space of policy updates. The resulting method, MeVa, is consistent and far-sighted, and does not require REINFORCE estimators. We analyze the behavior of our method on a toy game and compare to prior work on repeated matrix games.
Feature transformation aims to reconstruct an effective representation space by mathematically refining the existing features. It serves as a pivotal approach to combat the curse of dimensionality, enhance model generalization, mitigate data sparsity, and extend the applicability of classical models. Existing research predominantly focuses on domain knowledge-based feature engineering or learning latent representations. However, these methods, while insightful, lack full automation and fail to yield a traceable and optimal representation space. An indispensable question arises: Can we concurrently address these limitations when reconstructing a feature space for a machine-learning task? Our initial work took a pioneering step towards this challenge by introducing a novel self-optimizing framework. This framework leverages the power of three cascading reinforced agents to automatically select candidate features and operations for generating improved feature transformation combinations. Despite the impressive strides made, there was room for enhancing its effectiveness and generalization capability. In this extended journal version, we advance our initial work from two distinct yet interconnected perspectives: 1) We propose a refinement of the original framework, which integrates a graph-based state representation method to capture the feature interactions more effectively and develop different Q-learning strategies to alleviate Q-value overestimation further. 2) We utilize a new optimization technique (actor-critic) to train the entire self-optimizing framework in order to accelerate the model convergence and improve the feature transformation performance. Finally, to validate the improved effectiveness and generalization capability of our framework, we perform extensive experiments and conduct comprehensive analyses.
This paper investigates an interference-aware joint path planning and power allocation mechanism for a cellular-connected unmanned aerial vehicle (UAV) in a sparse suburban environment. The UAV's goal is to fly from an initial point and reach a destination point by moving along the cells to guarantee the required quality of service (QoS). In particular, the UAV aims to maximize its uplink throughput and minimize the level of interference to the ground user equipment (UEs) connected to the neighbor cellular BSs, considering the shortest path and flight resource limitation. Expert knowledge is used to experience the scenario and define the desired behavior for the sake of the agent (i.e., UAV) training. To solve the problem, an apprenticeship learning method is utilized via inverse reinforcement learning (IRL) based on both Q-learning and deep reinforcement learning (DRL). The performance of this method is compared to learning from a demonstration technique called behavioral cloning (BC) using a supervised learning approach. Simulation and numerical results show that the proposed approach can achieve expert-level performance. We also demonstrate that, unlike the BC technique, the performance of our proposed approach does not degrade in unseen situations.
Stock trading is one of the popular ways for financial management. However, the market and the environment of economy is unstable and usually not predictable. Furthermore, engaging in stock trading requires time and effort to analyze, create strategies, and make decisions. It would be convenient and effective if an agent could assist or even do the task of analyzing and modeling the past data and then generate a strategy for autonomous trading. Recently, reinforcement learning has been shown to be robust in various tasks that involve achieving a goal with a decision making strategy based on time-series data. In this project, we have developed a pipeline that simulates the stock trading environment and have trained an agent to automate the stock trading process with deep reinforcement learning methods, including deep Q-learning, deep SARSA, and the policy gradient method. We evaluate our platform during relatively good (before 2021) and bad (2021 - 2022) situations. The stocks we've evaluated on including Google, Apple, Tesla, Meta, Microsoft, and IBM. These stocks are among the popular ones, and the changes in trends are representative in terms of having good and bad situations. We showed that before 2021, the three reinforcement methods we have tried always provide promising profit returns with total annual rates around $70\%$ to $90\%$, while maintain a positive profit return after 2021 with total annual rates around 2% to 7%.
Offline reinforcement learning (RL) aims to learn policies from pre-existing datasets without further interactions, making it a challenging task. Q-learning algorithms struggle with extrapolation errors in offline settings, while supervised learning methods are constrained by model expressiveness. Recently, diffusion models (DMs) have shown promise in overcoming these limitations in single-agent learning, but their application in multi-agent scenarios remains unclear. Generating trajectories for each agent with independent DMs may impede coordination, while concatenating all agents' information can lead to low sample efficiency. Accordingly, we propose MADiff, which is realized with an attention-based diffusion model to model the complex coordination among behaviors of multiple agents. To our knowledge, MADiff is the first diffusion-based multi-agent learning framework, functioning as both a decentralized policy and a centralized controller. During decentralized executions, MADiff simultaneously performs teammate modeling, and the centralized controller can also be applied in multi-agent trajectory predictions. Our experiments demonstrate that MADiff outperforms baseline algorithms across various multi-agent learning tasks, highlighting its effectiveness in modeling complex multi-agent interactions. Our code is available at https://github.com/zbzhu99/madiff.
Offline reinforcement learning (RL) allows agents to learn effective, return-maximizing policies from a static dataset. Three popular algorithms for offline RL are Conservative Q-Learning (CQL), Behavior Cloning (BC), and Decision Transformer (DT), from the class of Q-Learning, Imitation Learning, and Sequence Modeling respectively. A key open question is: which algorithm is preferred under what conditions? We study this question empirically by exploring the performance of these algorithms across the commonly used D4RL and Robomimic benchmarks. We design targeted experiments to understand their behavior concerning data suboptimality, task complexity, and stochasticity. Our key findings are: (1) DT requires more data than CQL to learn competitive policies but is more robust; (2) DT is a substantially better choice than both CQL and BC in sparse-reward and low-quality data settings; (3) DT and BC are preferable as task horizon increases, or when data is obtained from human demonstrators; and (4) CQL excels in situations characterized by the combination of high stochasticity and low data quality. We also investigate architectural choices and scaling trends for DT on Atari and D4RL and make design/scaling recommendations. We find that scaling the amount of data for DT by 5x gives a 2.5x average score improvement on Atari.
Though robustness of networks to random attacks has been widely studied, intentional destruction by an intelligent agent is not tractable with previous methods. Here we devise a single-player game on a lattice that mimics the logic of an attacker attempting to destroy a network. The objective of the game is to disable all nodes in the fewest number of steps. We develop a reinforcement learning approach using deep Q-learning that is capable of learning to play this game successfully, and in so doing, to optimally attack a network. Because the learning algorithm is universal, we train agents on different definitions of robustness and compare the learned strategies. We find that superficially similar definitions of robustness induce different strategies in the trained agent, implying that optimally attacking or defending a network is sensitive the particular objective. Our method provides a new approach to understand network robustness, with potential applications to other discrete processes in disordered systems.
Many optimal control problems require the simultaneous output of discrete and continuous control variables. These problems are usually formulated as mixed-integer optimal control (MIOC) problems, which are challenging to solve due to the complexity of the solution space. Numerical methods such as branch-and-bound are computationally expensive and undesirable for real-time control. This paper proposes a novel hybrid-action reinforcement learning (HARL) algorithm, twin delayed deep deterministic actor-Q (TD3AQ), for MIOC problems. TD3AQ combines the advantages of both actor-critic and Q-learning methods, and can handle the discrete and continuous action spaces simultaneously. The proposed algorithm is evaluated on a plug-in hybrid electric vehicle (PHEV) energy management problem, where real-time control of the discrete variables, clutch engagement/disengagement and gear shift, and continuous variable, engine torque, is essential to maximize fuel economy while satisfying driving constraints. Simulation outcomes demonstrate that TD3AQ achieves control results close to optimality when compared with dynamic programming (DP), with just 4.69% difference. Furthermore, it surpasses the performance of baseline reinforcement learning algorithms.
Motion planning of autonomous agents in partially known environments with incomplete information is a challenging problem, particularly for complex tasks. This paper proposes a model-free reinforcement learning approach to address this problem. We formulate motion planning as a probabilistic-labeled partially observable Markov decision process (PL-POMDP) problem and use linear temporal logic (LTL) to express the complex task. The LTL formula is then converted to a limit-deterministic generalized B\"uchi automaton (LDGBA). The problem is redefined as finding an optimal policy on the product of PL-POMDP with LDGBA based on model-checking techniques to satisfy the complex task. We implement deep Q learning with long short-term memory (LSTM) to process the observation history and task recognition. Our contributions include the proposed method, the utilization of LTL and LDGBA, and the LSTM-enhanced deep Q learning. We demonstrate the applicability of the proposed method by conducting simulations in various environments, including grid worlds, a virtual office, and a multi-agent warehouse. The simulation results demonstrate that our proposed method effectively addresses environment, action, and observation uncertainties. This indicates its potential for real-world applications, including the control of unmanned aerial vehicles (UAVs).
Effective offline RL methods require properly handling out-of-distribution actions. Implicit Q-learning (IQL) addresses this by training a Q-function using only dataset actions through a modified Bellman backup. However, it is unclear which policy actually attains the values represented by this implicitly trained Q-function. In this paper, we reinterpret IQL as an actor-critic method by generalizing the critic objective and connecting it to a behavior-regularized implicit actor. This generalization shows how the induced actor balances reward maximization and divergence from the behavior policy, with the specific loss choice determining the nature of this tradeoff. Notably, this actor can exhibit complex and multimodal characteristics, suggesting issues with the conditional Gaussian actor fit with advantage weighted regression (AWR) used in prior methods. Instead, we propose using samples from a diffusion parameterized behavior policy and weights computed from the critic to then importance sampled our intended policy. We introduce Implicit Diffusion Q-learning (IDQL), combining our general IQL critic with the policy extraction method. IDQL maintains the ease of implementation of IQL while outperforming prior offline RL methods and demonstrating robustness to hyperparameters. Code is available at https://github.com/philippe-eecs/IDQL.
Quantum computing holds great potential for advancing the limitations of machine learning algorithms to handle higher dimensions of data and reduce overall training parameters in deep learning (DL) models. This study uses a trainable variational quantum circuit (VQC) on a gate-based quantum computing model to investigate the potential for quantum benefit in a model-free reinforcement learning problem. Through a comprehensive investigation and evaluation of the current model and capabilities of quantum computers, we designed and trained a novel hybrid quantum neural network based on the latest Qiskit and PyTorch framework. We compared its performance with a full-classical CNN with and without an incorporated VQC. Our research provides insights into the potential of deep quantum learning to solve a maze problem and, potentially, other reinforcement learning problems. We conclude that reinforcement learning problems can be practical with reasonable training epochs. Moreover, a comparative study of full-classical and hybrid quantum neural networks is discussed to understand these two approaches' performance, advantages, and disadvantages to deep-Q learning problems, especially on larger-scale maze problems larger than 4x4.
Despite advances in Reinforcement Learning, many sequential decision making tasks remain prohibitively expensive and impractical to learn. Recently, approaches that automatically generate reward functions from logical task specifications have been proposed to mitigate this issue; however, they scale poorly on long-horizon tasks (i.e., tasks where the agent needs to perform a series of correct actions to reach the goal state, considering future transitions while choosing an action). Employing a curriculum (a sequence of increasingly complex tasks) further improves the learning speed of the agent by sequencing intermediate tasks suited to the learning capacity of the agent. However, generating curricula from the logical specification still remains an unsolved problem. To this end, we propose AGCL, Automaton-guided Curriculum Learning, a novel method for automatically generating curricula for the target task in the form of Directed Acyclic Graphs (DAGs). AGCL encodes the specification in the form of a deterministic finite automaton (DFA), and then uses the DFA along with the Object-Oriented MDP (OOMDP) representation to generate a curriculum as a DAG, where the vertices correspond to tasks, and edges correspond to the direction of knowledge transfer. Experiments in gridworld and physics-based simulated robotics domains show that the curricula produced by AGCL achieve improved time-to-threshold performance on a complex sequential decision-making problem relative to state-of-the-art curriculum learning (e.g, teacher-student, self-play) and automaton-guided reinforcement learning baselines (e.g, Q-Learning for Reward Machines). Further, we demonstrate that AGCL performs well even in the presence of noise in the task's OOMDP description, and also when distractor objects are present that are not modeled in the logical specification of the tasks' objectives.
The graph colouring problem consists of assigning labels, or colours, to the vertices of a graph such that no two adjacent vertices share the same colour. In this work we investigate whether deep reinforcement learning can be used to discover a competitive construction heuristic for graph colouring. Our proposed approach, ReLCol, uses deep Q-learning together with a graph neural network for feature extraction, and employs a novel way of parameterising the graph that results in improved performance. Using standard benchmark graphs with varied topologies, we empirically evaluate the benefits and limitations of the heuristic learned by ReLCol relative to existing construction algorithms, and demonstrate that reinforcement learning is a promising direction for further research on the graph colouring problem.
Most offline reinforcement learning (RL) methods suffer from the trade-off between improving the policy to surpass the behavior policy and constraining the policy to limit the deviation from the behavior policy as computing $Q$-values using out-of-distribution (OOD) actions will suffer from errors due to distributional shift. The recently proposed \textit{In-sample Learning} paradigm (i.e., IQL), which improves the policy by quantile regression using only data samples, shows great promise because it learns an optimal policy without querying the value function of any unseen actions. However, it remains unclear how this type of method handles the distributional shift in learning the value function. In this work, we make a key finding that the in-sample learning paradigm arises under the \textit{Implicit Value Regularization} (IVR) framework. This gives a deeper understanding of why the in-sample learning paradigm works, i.e., it applies implicit value regularization to the policy. Based on the IVR framework, we further propose two practical algorithms, Sparse $Q$-learning (SQL) and Exponential $Q$-learning (EQL), which adopt the same value regularization used in existing works, but in a complete in-sample manner. Compared with IQL, we find that our algorithms introduce sparsity in learning the value function, making them more robust in noisy data regimes. We also verify the effectiveness of SQL and EQL on D4RL benchmark datasets and show the benefits of in-sample learning by comparing them with CQL in small data regimes.
Privacy-preserving vision must overcome the dual challenge of utility and privacy. Too much anonymity renders the images useless, but too little privacy does not protect sensitive data. We propose a novel design for privacy preservation, where the imagery is stored in quantum states. In the future, this will be enabled by quantum imaging cameras, and, currently, storing very low resolution imagery in quantum states is possible. Quantum state imagery has the advantage of being both private and non-private till the point of measurement. This occurs even when images are manipulated, since every quantum action is fully reversible. We propose a control algorithm, based on double deep Q-learning, to learn how to anonymize the image before measurement. After learning, the RL weights are fixed, and new attack neural networks are trained from scratch to break the system's privacy. Although all our results are in simulation, we demonstrate, with these first steps, that it is possible to control both privacy and utility in a quantum-based manner.