Multi-agent Reinforcement Learning Method Research Articles

Continuous self-adaptive optimization to learn multi-task multi-agent

Multi-agent reinforcement learning (MARL) methods have shown superior performance to solve a variety of real-world problems focusing on learning distinct policies for individual tasks. These approaches face problems when applied to the non-stationary real-world: agents trained in specialized tasks cannot achieve satisfied generalization performance across multiple tasks; agents have to learn and store specialized policies for individual task and reliable identities of tasks are hardly observable in practice. To address the challenge continuously adapting to multiple tasks in MARL, we formalize the problem into a two-stage curriculum. Single-task policies are learned with MARL approaches, after that we develop a gradient-based Self-Adaptive Meta-Learning algorithm, SAML, that cannot only distill single-task policies into a unified policy but also can facilitate the unified policy to continuously adapt to new incoming tasks. In addition, to validate the continuous adaptation performance on complex task, we extend the widely adopted StarCraft benchmark SMAC and develop a new multi-task multi-agent StarCraft environment, Meta-SMAC, for testing various aspects of continuous adaptation method. Our experiments with a population of agents show that our method enables significantly more efficient adaptation than reactive baselines across different scenarios.

Trajectory Optimization of CAVs in Freeway Work Zone considering Car-Following Behaviors Using Online Multiagent Reinforcement Learning

Work zone areas are frequent congested sections considered as the freeway bottleneck. Connected and autonomous vehicle (CAV) trajectory optimization can improve the operating efficiency in bottleneck areas by harmonizing vehicles’ manipulations. This study presents a joint trajectory optimization of cooperative lane changing, merging, and car-following actions for CAV control at a local merging point together with upstream points. The multiagent reinforcement learning (MARL) method is applied in this system, with one agent providing a merging advisory service at the merging point and controlling the inner-lane vehicles’ headway for smooth outer-lane vehicle merging, while other agents provide lane-changing advisory services at advance lane-changing points to control how vehicles make lane changes in advance and perform corresponding headway adjustment, similar to and jointly with the merging advisory service. Uniting all agents, the coordination graph (CG) method is applied to seek the global optimum, overcoming the exponential growth problem in MARL. Using MATLAB and the VISSIM COM interface, an online simulation platform is established. The simulation results show that MARL is effective for online computation with in-timing response. More importantly, comparisons of the results obtained in various scenarios demonstrate that the proposed system obtained smoother vehicle trajectories in all controlled sections, rather than only in the merging area, indicating that it can achieve better traffic conditions in freeway work zone areas.

Analysis of Evolutionary Dynamics for Bidding Strategy Driven by Multi-Agent Reinforcement Learning

In this letter, the evolutionary game theory (EGT) with replication dynamic equations (RDEs) is adopted to explicitly determine the factors affecting energy providers’ (EPs) willingness of using the market power to uplift the price in the bidding procedure, which could be simulated using the win-or-learn-fast policy hill climbing (WoLF-PHC) algorithm as a multi-agent reinforcement learning (MARL) method. Firstly, empirical and numerical connections between WoLF-PHC and RDEs is proved. Then, by formulating RDEs of the bidding procedure, three factors affecting the bidding strategy preference are revealed, including the load demand, severity of congestion, and the price cap. Finally, the impact of these factors on the converged bidding price is demonstrated in case studies, by simulating the bidding procedure driven by WoLF-PHC.

A new solution to distributed permutation flow shop scheduling problem based on NASH Q-Learning

Aiming at Distributed Permutation Flow-shop Scheduling Problems (DPFSPs), this study took the minimization of the maximum completion time of the workpieces to be processed in all production tasks as the goal, and took the multi-agent Reinforcement Learning (RL) method as the main frame of the solution model, then, combining with the NASH equilibrium theory and the RL method, it proposed a NASH Q-Learning algorithm for Distributed Flow-shop Scheduling Problem (DFSP) based on Mean Field (MF). In the RL part, this study designed a two-layer online learning mode in which the sample collection and the training improvement proceed alternately, the outer layer collects samples, when the collected samples meet the requirement of batch size, it enters to the inner layer loop, which uses the Q-learning model-free batch processing mode to proceed, and adopts neural network to approximate the value function to adapt to large-scale problems. By comparing the Average Relative Percentage Deviation (ARPD) index of the benchmark test questions, the calculation results of the proposed algorithm outperformed other similar algorithms, which proved the feasibility and efficiency of the proposed algorithm.

Multi-target tracking for unmanned aerial vehicle swarms using deep reinforcement learning

In recent years, deep reinforcement learning (DRL) has proved its great potential in multi-agent cooperation. However, how to apply DRL to multi-target tracking (MTT) problem for unmanned aerial vehicle (UAV) swarms is challenging: 1) the scale of UAVs may be large, but the existing multi-agent reinforcement learning (MARL) methods that rely on global or joint information of all agents suffer from the dimensionality curse; 2) the dimension of each UAV’s received information is variable, which is incompatible with the neural networks with fixed input dimensions; 3) the UAVs are homogeneous and interchangeable that each UAV’s policy should be irrelevant to the permutation of its received information. To this end, we propose a DRL method for UAV swarms to solve the MTT problem. Firstly, a decentralized swarm-oriented Markov Decision Process (MDP) model is presented for UAV swarms, which involves each UAV’s local communication and partial observation. Secondly, to achieve better scalability, a cartogram feature representation (FR) is proposed to integrate the variable-dimensional information set into a fixed-shape input variable, and the cartogram FR can also maintain the permutation irrelevance to the information. Then, the double deep Q-learning network with dueling architecture is adapted to the MTT problem, and the experience-sharing training mechanism is adopted to learn the shared cooperative policy for UAV swarms. Extensive experiments are provided and the results show that our method can successfully learn a cooperative tracking policy for UAV swarms and outperforms the baseline method in the tracking ratio and scalability.

Towards Self-X cognitive manufacturing network: An industrial knowledge graph-based multi-agent reinforcement learning approach

Empowered by the advanced cognitive computing, industrial Internet-of-Things, and data analytics techniques, today’s smart manufacturing systems are ever-increasingly equipped with cognitive capabilities, towards an emerging Self-X cognitive manufacturing network with higher level of automation. Nevertheless, to our best knowledge, the readiness of ‘Self-X’ levels (e.g., self-configuration, self-optimization, and self-adjust/adaptive/healing) is still in the infant stage. To pave its way, this work stepwise introduces an industrial knowledge graph (IKG)-based multi-agent reinforcement learning (MARL) method for achieving the Self-X cognitive manufacturing network. Firstly, an IKG should be formulated based on the extracted empirical knowledge and recognized patterns in the manufacturing process, by exploiting the massive human-generated and machine-sensed multimodal data. Then, a proposed graph neural network-based embedding algorithm can be performed based on a comprehensive understanding of the established IKG, to achieve semantic-based self-configurable solution searching and task decomposition. Moreover, a MARL-enabled decentralized system is presented to self-optimize the manufacturing process, and to further complement the IKG towards Self-X cognitive manufacturing network. An illustrative example of multi-robot reaching task is conducted lastly to validate the feasibility of the proposed approach. As an explorative study, limitations and future perspectives are also highlighted to attract more open discussions and in-depth research for ever smarter manufacturing.

Autonomous Bus Fleet Control Using Multiagent Reinforcement Learning

Autonomous buses are becoming increasingly popular and have been widely developed in many countries. However, autonomous buses must learn to navigate the city efficiently to be integrated into public transport systems. Efficient operation of these buses can be achieved by intelligent agents through reinforcement learning. In this study, we investigate the autonomous bus fleet control problem, which appears noisy to the agents owing to random arrivals and incomplete observation of the environment. We propose a multi-agent reinforcement learning method combined with an advanced policy gradient algorithm for this large-scale dynamic optimization problem. An agent-based simulation platform was developed to model the dynamic system of a fixed stop/station loop route, autonomous bus fleet, and passengers. This platform was also applied to assess the performance of the proposed algorithm. The experimental results indicate that the developed algorithm outperforms other reinforcement learning methods in the multi-agent domain. The simulation results also reveal the effectiveness of our proposed algorithm in outperforming the existing scheduled bus system in terms of the bus fleet size and passenger wait times for bus routes with comparatively lesser number of passengers.

Optimizing task scheduling in human-robot collaboration with deep multi-agent reinforcement learning

Human-Robot Collaboration (HRC) presents an opportunity to improve the efficiency of manufacturing processes. However, the existing task planning approaches for HRC are still limited in many ways, e.g., co-robot encoding must rely on experts’ knowledge and the real-time task scheduling is applicable within small state-action spaces or simplified problem settings. In this paper, the HRC assembly working process is formatted into a novel chessboard setting, in which the selection of chess piece move is used to analogize to the decision making by both humans and robots in the HRC assembly working process. To optimize the completion time, a Markov game model is considered, which takes the task structure and the agent status as the state input and the overall completion time as the reward. Without experts’ knowledge, this game model is capable of seeking for correlated equilibrium policy among agents with convergency in making real-time decisions facing a dynamic environment. To improve the efficiency in finding an optimal policy of the task scheduling, a deep-Q-network (DQN) based multi-agent reinforcement learning (MARL) method is applied and compared with the Nash-Q learning, dynamic programming and the DQN-based single-agent reinforcement learning method. A height-adjustable desk assembly is used as a case study to demonstrate the effectiveness of the proposed algorithm with different number of tasks and agents.

A Multi-agent Reinforcement Learning Method for Role Differentiation Using State Space Filters with Fluctuation Parameters

Recently, there have been many studies on Multi-agent Reinforcement Learning (MARL) in which each autonomous agent obtains its own control rule by RL. Here, we hypothesize that different agents having individuality is more effective than uniform agents in terms of role differentiation in MARL. We have previously proposed a promoting method of role differentiation using a waveform changing parameter in MARL. In this paper, we confirm the effectiveness of role differentiation by introducing the waveform changing parameter into a state space filter through computational examples using “Pursuit Game” as a multi-agent task.

Coordination Between Individual Agents in Multi-Agent Reinforcement Learning

The existing multi-agent reinforcement learning methods (MARL) for determining the coordination between agents focus on either global-level or neighborhood-level coordination between agents. However the problem of coordination between individual agents is remain to be solved. It is crucial for learning an optimal coordinated policy in unknown multi-agent environments to analyze the agent's roles and the correlation between individual agents. To this end, in this paper we propose an agent-level coordination based MARL method. Specifically, it includes two parts in our method. The first is correlation analysis between individual agents based on the Pearson, Spearman, and Kendall correlation coefficients; And the second is an agent-level coordinated training framework where the communication message between weakly correlated agents is dropped out, and a correlation based reward function is built. The proposed method is verified in four mixed cooperative-competitive environments. The experimental results show that the proposed method outperforms the state-of-the-art MARL methods and can measure the correlation between individual agents accurately.

A Deep Ensemble Method for Multi-Agent Reinforcement Learning: A Case Study on Air Traffic Control

Reinforcement learning (RL), a promising framework for data-driven decision making in an uncertain environment, has successfully been applied in many real-world operation and control problems. However, the application of RL in a large-scale decentralized multi-agent environment remains a challenging problem due to the partial observability and limited communications between agents. In this paper, we develop a model-based kernel RL approach and a model-free deep RL approach for learning a decentralized, shared policy among homogeneous agents. By leveraging the strengths of both these methods, we further propose a novel deep ensemble multi-agent reinforcement learning (MARL) method that efficiently learns to arbitrate between the decisions of the local kernel-based RL model and the wider-reaching deep RL model. We validate the proposed deep ensemble method on a highly challenging real-world air traffic control problem, where the goal is to provide effective guidance to aircraft to avoid air traffic congestion, conflicting situations, and to improve arrival timeliness, by dynamically recommending adjustments of aircraft speeds in real-time. Extensive empirical results from an open-source air traffic management simulation model, developed by Eurocontrol and built on a real-world data set including thousands of aircrafts, demonstrate that our proposed deep ensemble MARL method significantly outperforms three state-of-the-art benchmark approaches.

Resource Allocation for Delay-Sensitive Vehicle-to-Multi-Edges (V2Es) Communications in Vehicular Networks: A Multi-Agent Deep Reinforcement Learning Approach

The rapid development of internet of vehicles (IoV) has recently led to the emergence of diverse intelligent vehicular applications such as automatic driving, auto navigation, and advanced driver assistance, etc. However, the current vehicular communication framework, such as vehicle-to-vehicle (V2V), vehicle-to-cloud (V2C), and vehicle-to-roadside infrastructure (V2I), still remain challenging in supporting these intelligent and delay-sensitive applications, due to its long communication latency or low computational capability. Besides that, the traditional vehicle network is prone to be unavailable because of the mobility with high-speed of the vehicles. To address these issues, this paper proposes a vehicle-to-multi-edges (V2Es) communication framework in vehicular networks. By utilizing the resource of edge nodes in the proximity, the emergency information or services of vehicles can be timely processed and completed, which improves the service quality of vehicles. Furthermore, we define a joint task offloading and edge caching problem, targeting optimizing both the latency of services and energy consumption of vehicles. Based on this, we propose a multi-agent reinforcement learning (RL) method to learn the dynamic communication status between vehicles and edge nodes, and make decisions on task offloading and edge caching. Finally, results of the simulation show that our proposal is able to learn the scheduling policy more quickly and effectively, and reduce the service latency by more than 10% on average.

Network-wide traffic signal control optimization using a multi-agent deep reinforcement learning

Inefficient traffic control may cause numerous problems such as traffic congestion and energy waste. This paper proposes a novel multi-agent reinforcement learning method, named KS-DDPG (Knowledge Sharing Deep Deterministic Policy Gradient) to achieve optimal control by enhancing the cooperation between traffic signals. By introducing the knowledge-sharing enabled communication protocol, each agent can access to the collective representation of the traffic environment collected by all agents. The proposed method is evaluated through two experiments respectively using synthetic and real-world datasets. The comparison with state-of-the-art reinforcement learning-based and conventional transportation methods demonstrate the proposed KS-DDPG has significant efficiency in controlling large-scale transportation networks and coping with fluctuations in traffic flow. In addition, the introduced communication mechanism has also been proven to speed up the convergence of the model without significantly increasing the computational burden.

Hierarchical multiagent reinforcement learning schemes for air traffic management

In this work we investigate the use of hierarchical multiagent reinforcement learning methods for the computation of policies to resolve congestion problems in the air traffic management domain. To address cases where the demand of airspace use exceeds capacity, we consider agents representing flights, who need to decide on ground delays at the pre-tactical stage of operations, towards executing their trajectories while adhering to airspace capacity constraints. Hierarchical reinforcement learning manages to handle real-world problems with high complexity, by partitioning the task into hierarchies of states and/or actions. This provides an efficient way of exploring the state–action space and constructing an advantageous decision-making mechanism. We first establish a general framework of hierarchical multiagent reinforcement learning, and then, we further formulate four alternative schemes of abstractions, on states, actions, or both. To quantitatively assess the quality of solutions of the proposed approaches and show the potential of the hierarchical methods in resolving the demand–capacity balance problem, we provide experimental results on real-world evaluation cases, where we measure the average delay per flight and the number of flights with delays.

A Collaborative Multiagent Reinforcement Learning Method Based on Policy Gradient Potential.

Gradient-based method has been extensively used in today's multiagent reinforcement learning (MARL). In a gradient-based MARL algorithm, each agent updates its parameterized strategy in the direction of the gradient of some performance index. However, studies on the convergence of the existing gradient-based MARL algorithms for identical interest games are quite few. In this article, we propose a policy gradient potential (PGP) algorithm that takes PGP as the source of information for guiding the strategy update, as opposed to the gradient itself, to learn the optimal joint strategy that has a maximal global reward. Since the payoff matrix and the joint strategy are often unavailable to the learning agents in reality, we consider the probability of obtaining the maximal reward as the performance index. Theoretical analysis of the PGP algorithm on the continuous model involving an identical interest repeated game shows that if the component action of every optimal joint action is unique, the critical points corresponding to all optimal joint actions are asymptotically stable. The PGP algorithm is experimentally studied and compared against other MARL algorithms on two commonly used collaborative tasks-the robots leaving a room task and the distributed sensor network task, as well as a real-world minefield navigation problem where only local state and local reward information are available. The results show that the PGP algorithm outperforms the other algorithms in terms of the cumulative reward and the number of time steps used in an episode.

A Cooperative Multi-Agent Reinforcement Learning Method Based on Coordination Degree

Multi-agent reinforcement learning (MARL) has become a prevalent method for solving cooperative problems owing to its tractable implementation and task distribution. The goal of the MARL algorithms for fully cooperative scenarios is to obtain the optimal joint strategy that maximizes the expected common cumulative reward for all agents. However, to date, the analysis of MARL dynamics has focused on repeated games with few agents and actions. To this end, we propose a cooperative MARL algorithm based on the coordination degree (CMARL-CD) and analyze its dynamics in more general cases in which repeated games with more agents and actions are considered. Theoretical analysis shows that if the component action of every optimal joint action is unique, all optimal joint actions are asymptotically stable critical points. The CMARL-CD algorithm realizes coordination among agents without the need to estimate the global Q-value function. Each agent estimates the coordination degree of its own action, which represents the potential of being the optimal action. The efficacy of the CMARL-CD algorithm is studied through repeated games and stochastic games.

Multi-vehicle routing problems with soft time windows: A multi-agent reinforcement learning approach

Multi-vehicle routing problem with soft time windows (MVRPSTW) is an indispensable constituent in urban logistics distribution systems. Over the past decade, numerous methods for MVRPSTW have been proposed, but most are based on heuristic rules that require a large amount of computation time. With the current rapid increase of logistics demands, traditional methods incur the dilemma between computational efficiency and solution quality. To efficiently solve the problem, we propose a novel reinforcement learning algorithm called the Multi-Agent Attention Model that can solve routing problem instantly benefit from lengthy offline training. Specifically, the vehicle routing problem is regarded as a vehicle tour generation process, and an encoder-decoder framework with attention layers is proposed to generate tours of multiple vehicles iteratively. Furthermore, a multi-agent reinforcement learning method with an unsupervised auxiliary network is developed for the model training. By evaluated on four synthetic networks with different scales, the results demonstrate that the proposed method consistently outperforms Google OR-Tools and traditional methods with little computation time. In addition, we validate the robustness of the well-trained model by varying the number of customers and the capacities of vehicles.

Learning to Delay in Ride-Sourcing Systems: A Multi-Agent Deep Reinforcement Learning Framework

Ride-sourcing services are now reshaping the way people travel by effectively connecting drivers and passengers through mobile internets. Online matching between idle drivers and waiting passengers is one of the most key components in a ride-sourcing system. The average pickup distance or time is an important measurement of system efficiency since it affects both passengers’ waiting time and drivers’ utilization rate. It is naturally expected that a more effective bipartite matching (with smaller average pickup time) can be implemented if the platform accumulates more idle drivers and waiting passengers in the matching pool. A specific passenger request can also benefit from a delayed matching since he/she may be matched with closer idle drivers after waiting for a few seconds. Motivated by the potential benefits of delayed matching, this paper establishes a two-stage framework which incorporates a combinatorial optimization and multi-agent deep reinforcement learning methods. The multi-agent reinforcement learning methods are used to dynamically determine the delayed time for each passenger request (or the time at which each request enters the matching pool), while the combinatorial optimization conducts an optimal bipartite matching between idle drivers and waiting passengers in the matching pool. Four tailored reinforcement learning methods, delayed multi-agent deep Q learning (Delayed-M-DQN), delayed multi-agent actor-critic (Delayed-M-A2C), delayed multi-agent Proximal Policy Optimization (Delayed-M-PPO), and delayed multi-agent actor-critic with experience replay (Delayed-M-ACER), are developed. Through extensive empirical experiments with a well-designed simulator, we show that the proposed framework is able to remarkably improve system performances, by well balancing the trade-off among pick-up time, matching time, successful matching rate.

Bi-Level Actor-Critic for Multi-Agent Coordination

Coordination is one of the essential problems in multi-agent systems. Typically multi-agent reinforcement learning (MARL) methods treat agents equally and the goal is to solve the Markov game to an arbitrary Nash equilibrium (NE) when multiple equilibra exist, thus lacking a solution for NE selection. In this paper, we treat agents unequally and consider Stackelberg equilibrium as a potentially better convergence point than Nash equilibrium in terms of Pareto superiority, especially in cooperative environments. Under Markov games, we formally define the bi-level reinforcement learning problem in finding Stackelberg equilibrium. We propose a novel bi-level actor-critic learning method that allows agents to have different knowledge base (thus intelligent), while their actions still can be executed simultaneously and distributedly. The convergence proof is given, while the resulting learning algorithm is tested against the state of the arts. We found that the proposed bi-level actor-critic algorithm successfully converged to the Stackelberg equilibria in matrix games and find a asymmetric solution in a highway merge environment.

Optimal Exploration Algorithm of Multi-Agent Reinforcement Learning Methods (Student Abstract)

Exploration efficiency challenges for multi-agent reinforcement learning (MARL), as the policy learned by confederate MARL depends on the interaction among agents. Less informative reward also restricts the learning speed of MARL in comparison with the informative label in supervised learning. This paper proposes a novel communication method which helps agents focus on different exploration subarea to guide MARL to accelerate exploration. We propose a predictive network to forecast the reward of current state-action pair and use the guidance learned by the predictive network to modify the reward function. An improved prioritized experience replay is employed to help agents better take advantage of the different knowledge learned by different agents. Experimental results demonstrate that the proposed algorithm outperforms existing methods in cooperative multi-agent environments.