Multi-agent Simulation Research Articles

The Effect of Platform Data Quality on Tax Compliance in Digital Economy: A Multiagent Based Simulation

The expeditious development of the digital economy has posed critical challenges for tax compliance. Recent reforms and technological changes, such as the emergency of platform data sharing, perform as potential instruments to potentially solve the tax compliance issue in the digital economy. This article innovatively combines tax compliance and digital economy taxation and proposes an expected utility theory model and a multiagent simulation model of tax compliance in the context of the digital economy, incorporating an important impact factor—platform data quality. In addition, the discrepancy and applicable conditions of three audit modes have been first compared and analyzed in this research: unconstrained signal audit, constrained random audit, and constrained signal audit. Applying computational experiments, this paper further investigates the effect of platform data quality on tax compliance in the digital economy. The key findings include: (1) the signal role of platform data is affected by three factors—the quality of platform data, the audit intensity and audit rate; (2) the signal role of platform data can be negative in given situations; and (3) the tax compliance rate varies in three different audit modes, which depend on the platform data quality, audit rate, and audit intensity. Based on the findings, this article provides policy recommendations on tax collection and management in digital economy.

Decentralized policy learning with partial observation and mechanical constraints for multiperson modeling

Extracting the rules of real-world multi-agent behaviors is a current challenge in various scientific and engineering fields. Biological agents independently have limited observation and mechanical constraints; however, most of the conventional data-driven models ignore such assumptions, resulting in lack of biological plausibility and model interpretability for behavioral analyses. Here we propose sequential generative models with partial observation and mechanical constraints in a decentralized manner, which can model agents’ cognition and body dynamics, and predict biologically plausible behaviors. We formulate this as a decentralized multi-agent imitation-learning problem, leveraging binary partial observation and decentralized policy models based on hierarchical variational recurrent neural networks with physical and biomechanical penalties. Using real-world basketball and soccer datasets, we show the effectiveness of our method in terms of the constraint violations, long-term trajectory prediction, and partial observation. Our approach can be used as a multi-agent simulator to generate realistic trajectories using real-world data.

Integration of Level 3 probabilistic risk assessment for nuclear power plants with transportation simulation considering earthquake hazards

This research improves the realism of Level 3 probabilistic risk assessment (PRA) for nuclear power plants (NPPs) to avoid subjective expert judgement by modeling the evacuation behavior of the residents and making two advancements. First, evacuation speed taken from a transportation simulation code was used as an input to the Level 3 PRA code, i.e., the MELCOR Accident Consequence Code System (MACCS). The transportation simulation code uses the Multiagent Transport Simulation (MATSim) evacuation extension software. MATSim is interfaced with MACCS, by generating input-output communication through evacuation routes and speed distributions. Integrating the agent-based transportation simulation with the Level 3 PRA code enables the explicit incorporation of spatiotemporal evacuation processes into NPP offsite risk assessment. Second, this research is the first to explicitly incorporate the probability of seismic damage to roadway bridges into Level 3 PRA of NPPs. The likelihood of seismic bridge damage, given an earthquake, is estimated using the HAZUS program developed by the Federal Emergency Management Agency (FEMA). The road closure due to seismic damage, estimated by HAZUS, is used as input to the agent-based transportation simulation, where evacuation routes and speed distributions are calculated considering the impact of seismic damage. These advancements can improve the realism of Level 3 PRA for NPPs by enhancing the evacuation analysis resolution and reducing the reliance on subjective judgment to characterize evacuation routes, speed, and seismic impact on Level 3 PRA. Integrating the transportation simulation and the natural hazard risk assessment with the Level 3 PRA code can also provide more in-depth causal insights for offsite risk management, such as recommending optimal evacuation routes in an emergency response plan. These contributions are demonstrated using the Sequoyah NPP case study adopted from the state-of-the-art reactor consequence analyses (SOARCA) conducted by the U.S. Nuclear Regulatory Commission (NRC).

Neurodynamic approaches for multi-agent distributed optimization

This paper considers a class of multi-agent distributed convex optimization with a common set of constraints and provides several continuous-time neurodynamic approaches. In problem transformation, l1 and l2 penalty methods are used respectively to cast the linear consensus constraint into the objective function, which avoids introducing auxiliary variables and only involves information exchange among primal variables in the process of solving the problem. For nonsmooth cost functions, two differential inclusions with projection operator are proposed. Without convexity of the differential inclusions, the asymptotic behavior and convergence properties are explored. For smooth cost functions, by harnessing the smoothness of l2 penalty function, finite- and fixed-time convergent algorithms are provided via a specifically designed average consensus estimator. Finally, several numerical examples in the multi-agent simulation environment are conducted to illustrate the effectiveness of the proposed neurodynamic approaches.

Comparison of decentralised fast-charging strategies for long-distance trips with electric vehicles

Despite noticeable improvements in electric vehicle (EV) technology, EV charging time remains high, and the lack of fast-charging infrastructure extends the travelling time for long-distance trips even further. Charging strategies to coordinate EVs’ charging plans dynamically will help reduce station waiting times. This paper aims to quantify and compare the performances of two decentralised strategies using regular information-sharing between EVs and stations to estimate future waiting time in charging stations accurately. The communication strategy relies only on the communication of optimal charging plan computed by EVs to stations without any reservation. In contrast, the reservation strategy enables additionally the booking of optimal charging time slots. In a dedicated multi-agent simulation framework, we tested both strategies on different levels of infrastructure saturation. The communication strategy shows the best performance especially when combined with more frequent charging stops.

A multi-agent simulation based train platforming research for facilitating passenger transfer in a high-speed railway station

The train platforming problem is one of the most investigated topics in the train operation field. However, a systematic understanding of how train platforming improves the convenience of passenger transfer procedures is still lacking. To face the rapidly growing transfer demands in the high-speed railway station, this paper proposes a transfer-facilitating-oriented train platforming approach based on a multi-agent train operation simulation model. By evaluating the transfer connection quality using an indicator named Transfer Connection Strength(TCS), which considers transfer distance and transfer time, the proposed transfer-facilitating-oriented train platforming strategies can establish more convenient transfer connections and so improve the transfer service. A track occupation horizon based on the modelling of accurate train running processes is utilized to detect route conflicts and guarantee conflict-free train operations in the simulation model, enabling the solutions’ feasibility of train platforming to be taken into account. A series of simulation experiments are conducted using real data of Zhengzhou East station in China. The experiment results show that, the proposed transfer-facilitating-oriented train platforming strategies are capable of obtaining a greater number of convenient transfer connections. The feasibility can be guaranteed by permitted train delays and route conflict detection for the succeeding trains in a forward-looking period. Additionally, the platform assignment plans can be generated with a specified priority of transfer distance or transfer time.

A dynamic and reactive routing protocol for the physical internet network

The Physical Internet (PI) offers an innovative approach to logistics networks, focussing on operational, digital, and physical interconnectivity through encapsulation, interfaces, and protocols. However, PI network can be complex to manage due to its dynamic structure and the need for routing protocols that adapt to continually evolving circumstances. This paper proposes a dynamic and reactive routing protocol for a PI sub-network, leveraging well-known Internet techniques, namely ‘Route Request’ and ‘Route Reply’. The objective is to enable dynamic, self-starting, multi-hop routing among nodes, ensuring continuous connectivity even amidst network disruptions. Moreover, the protocol incorporates reactive assignment algorithms to address disruptions, such as product shortages and resource unavailability. Additionally, the protocol is innovative in considering each container and resource's preferences and local knowledge of the system's known state. To include preferences in decision-making, we incorporate the Technique for Order Preference by Similarity to Ideal Solution and goal programming. Finally, we use multi-agent simulation and real order data from two major French retail chains to evaluate the protocol's performance. Results showed improved routing robustness and efficiency amidst disruptions compared to state-of-the-art methods. Various scenarios have been examined to yield promising insights for future research and practical applications in the realm of the PI.

Multiagent Simulation of Seasonal Influenza from Preinfectious People in Closed Spaces

This study proposes a discrete mathematical Susceptible–Exposed–Preinfectious–Infectious–Recovered (SEPIR) states model for seasonal influenza. In a previous study, focusing on infections by preinfectious people using preexisting data, the author showed that the super-spreading of seasonal influenza occurred before the day that the first patients were discovered (D-day). In addition, when people do not take precautionary measures, the infectivity rate (from preinfected people) was determined as 0.041. After D-day in the community, the implementation of countermeasures was observed to reduce the infectivity rate to 0.002 and 0.013 in working and living spaces, respectively. The number of infectious people can be estimated by summing up each group in the community. This study performed a multiagent simulation (MAS) of seasonal influenza from preinfectious people in closed spaces based on decomposability. Then, the basic simulation is validated and the appropriateness of infective rates, changed infectivity rates, and near decomposability is confirmed.

Multi-agent simulation for multi-mode travel policy to improve park and ride efficiency

The efficiency of the existing park and ride transfer system is often restricted by the supply level of public transport and parking shortage. We aim to analyze multiple policy scenarios such as public transport supply and parking reservation through simulation to provide operable suggestions for urban traffic managers. So, an agent-based model (ABM) is established by us to simulate a multi-mode network that includes roads, buses, metro, and walking. Through this network, the whole travel process of travellers’ parking, transferring to public transport, and then walking to their destination can be simulated. Taking Suzhou Guanqian Street Commercial District as a case shows that compared with reducing the existing public transport fares, reducing the waiting time can attract car travellers to choose park and ride. Under the coexistence of reserved and non-reserved parking vehicles, with the increase in the proportion of reservations, the system can reduce cruising-for-parking and improve the traffic environment in the downtown area. In addition, the combination strategy simulation is employed to investigate the effects of various strategies on the urban traffic system. This approach offers a basis for policymakers to formulate informed decisions.

MAES: a ROS 2-compatible simulation tool for exploration and coverage algorithms

With the aim of allowing the efficient and realistic simulation of swarm algorithms for exploration and coverage, we present the tool Multi-Agent Exploration Simulator (MAES), which is an open-source physics-based discrete step multi-robot simulator. MAES features movement in a continuous 2D space, realistic physics based on the Unity framework, advanced visualization techniques such as heatmaps, custom wireless signal degradation, both randomly generated and custom user-provided maps, and a ROS (Robot Operating System) interface. This latter characteristic could allow to port the simulated algorithms to real-world robots. We present performance tests, conducted with rather modest hardware, showing that MAES is able to simulate up to 5 robots in ROSMode (using the ROS integration) and up to 120 robots in UnityMode (development performed directly into the C# Unity Editor). A usability test was conducted which hinted that the target audience of robotics researchers and developers is able to quickly install, setup, and use MAES for implementing simple robot logic.

The influencing factors underlying the vicious ecological vulnerability-low productivity-poverty cycle in China and overcoming its related policies

There is often a vicious cycle that occurs in China encompassing ecological vulnerability, low productivity, and poverty. Existing research has not applied multiagent modelling and simulation (MAMS), which is a method suitable for analysing such complex systems. Therefore, the MAMS is here used to explore potential strategies for breaking this cycle. The MAMS method is based on complex adaptive systems and computer programs, and it includes both theoretical and simulation models, which can be used to simulate different scenarios and obtain visualized results. To sample representative poverty-stricken areas in China, the authors designed five breakthrough policy scenarios. The simulation results of these scenarios indicate that increasing the amount of arable land decreases the number of poor people. However, increasing the direct interventions of government does not reduce the number of the poor, nor does it change the Gini coefficient. On the other hand, increasing the number and variety of poverty alleviation opportunities available to the poor leads to a decrease in both the number of poor people and the Gini coefficient. These results of our five scenarios indicate that the optimal policy portfolio could be obtained by increasing the amount of arable land and providing more varied opportunities to help the poor participate in market activities while reducing direct government intervention. The combined design of these policies is conducive to breaking the vicious cycle of poverty.

How convenience and informational tools shape waste separation behavior: A social network approach

Geopolitical risks such as energy crises and local or regional instability aggravate the growing conflict between limited natural resources and increasing consumption, making waste management—using waste as a resource—one of the most urgent agendas in the sustainability domain. Considering individuals' dual attributes (i.e., attitudes and behaviors), we develop an interaction model to investigate waste separation behavior that accounts for convenience and informational tools. Multiagent simulations are performed using data collected from 1056 residents in the top ten Chinese cities by volume of household waste. The results indicate that convenience tools can induce residents to separate waste. In particular, a 160% increase in the adoption rate is observed in response to decreasing (increasing) the distance to collection sites (accessibility to temporary storage facilities for recyclables) from 10 (0) to 0 (10). Publicity and education promote waste separation, as approximately 72% of individuals choose to separate waste when increasing the intensity of popular science propaganda from 0 to 0.2. Interestingly, social interaction appears to have a two-sided effect on waste separation behavior. The adoption rate of such behavior increases with the number of one's connections in sparse social networks, whereas the opposite is found in close social networks. The findings reveal the differentiated impacts of convenience and informational tools on waste minimization and provide new ideas for formulating customized strategies to promote waste separation in sparse (increasing individuals' number of connections) or close (monitoring rumors to achieve a tightly controlled intensity of interaction) social network scenarios.

Digital Twinning and Optimization of Manufacturing Process Flows

Abstract The new wave of Industry 4.0 is transforming manufacturing factories into data-rich environments. This provides an unprecedented opportunity to feed large amounts of sensing data collected from the physical factory into the construction of digital twin (DT) in cyberspace. However, little has been done to fully utilize the DT technology to improve the smartness and autonomous levels of small and medium-sized manufacturing factories. Indeed, only a small fraction of small and medium-sized manufacturers (SMMs) has considered implementing DT technology. There is an urgent need to exploit the full potential of data analytics and simulation-enabled DTs for advanced manufacturing. Hence, this paper presents the design and development of DT models for simulation optimization of manufacturing process flows. First, we develop a multi-agent simulation model that describes nonlinear and stochastic dynamics among a network of interactive manufacturing things, including customers, machines, automated guided vehicles (AGVs), queues, and jobs. Second, we propose a statistical metamodeling approach to design sequential computer experiments to optimize the utilization of AGV under uncertainty. Third, we construct two new graph models—job flow graph and AGV traveling graph—to track and monitor the real-time performance of manufacturing jobshops. The proposed simulation-enabled DT approach is evaluated and validated with experimental studies for the representation of a real-world manufacturing factory. Experimental results show that the proposed methodology effectively transforms a manufacturing jobshop into a new generation of DT-enabled smart factories. The sequential design of experiments effectively reduces the computation overhead of expensive simulations while optimally scheduling the AGV to achieve production throughput cost-effectively. This research is strongly promised to help SMMs fully utilize big data and DT technology for gaining competitive advantages in the global marketplace.

Exploring the theoretical and practical implications of eigenbehavior at the intersection of second-order cybernetics and ecosystem management

PurposeIn this conceptual paper the authors aim to discuss the implications of a second-order cybernetic approach and eigenbehavior in ecosystem management. The authors argue that traditional management practices rely on the agreement of stakeholders, but this approach cannot satisfy multiple-objective realities, making radical constructivism necessary.Design/methodology/approachThe authors examined this conjecture using a multi-agent simulation exercise. Agents display individual behavior consistent with the organizing principles proposed by von Foerster (eigenform and eigenbehavior), Hoffman (perception and fitness) and Maturana (domains of realization). This exercise is complemented with an ecosystem management classification exercise to show how convergence is achieved when a team of trained professionals describes entities of low versus high complexity in a spatially explicit domain.FindingsThe authors showed that eigenforms can diverge significantly depending on the complexity of the problem. The capacity of observers to stabilize an eigenform and create a common ground for understanding depends on the complexity of the problem at hand and their cognitive diversity. The authors highlighted the difficulties that arose when observation and modification through the intervention of the environment cannot be detached. The authors argue that this situation deeply permeates ecosystem resource management due to the unpredictable outcomes displayed by entities embedded into open system dynamics.Originality/valueFinally, the authors propose that the association between indication and transformation can be operationally interconnected as Eigenperception, which corresponds to a novel dynamic state bridging the gap between noise and eigenform. This concept of Eigenperception encapsulates entities that arise from eigenbehavior, encompassing observations, meanings, human transformative actions and ecological processes, all in a simultaneous manner.

Research on the evolution law of cloud manufacturing service ecosystem based on multi-agent behavior simulation

Monitoring and evaluating the health of the cloud manufacturing service ecosystem (CMSE) is critical to ensuring the long-term development of the cloud manufacturing service platform. The behavior patterns of three types of market entities, service demander (SD), service provider (SP) and platform operator (PO), have an important impact on the evolution trend of CMSE. The formulation of platform transaction rules and development of operation and regulation strategies need clarified the evolution law of the CMSE. Therefore, the evolution framework of the CMSE is constructed, the behavior modes of market entities such as SP, SD, and PO are established respectively, and multi-agent behavior simulation experiments are carried out. Simulation analysis showed that the more sensitive the SP was to the transaction activity in the system, the earlier the ecosystem in which it was located enters the stable period. When the sensitivity k of SD to the number of SPs that are available for required services in the system was between 0.45 and 0.5, the evolution trend of CMSE would show an obvious turning point from rapid growth to shrinkage. PO could adopt flexible charging strategies at different stages of the evolution of the CMSE to maximize revenue.

Examining the synergistic diffusion process of carbon capture and renewable energy generation technologies under market environment: A multi-agent simulation analysis

To achieve carbon neutrality, the power industry will vigorously promote renewable energy generation and carbon capture and storage (CCS) technologies. The diffusion of these technologies will vary depending on the incentive policies and market environments. Considering the progressive development of China's electricity market reform, this study constructs a multi-agent simulation model for the long-term low-carbon transformation of power systems, simulating the effects of different market environment and policy tools on the synergistic diffusion process of CCS and renewable energy generation technologies. According to the research results, when the price constraints in the electricity market are relaxed, opting for a differentiated clearing price demonstrates greater benefits for the diffusion of CCS technology and the reduction of carbon emissions compared to using a uniform clearing price. In this situation, the spread of renewable energy generation technologies aids in lowering the average market electricity price, thereby offsetting any potential price hikes resulting from the expansion of CCS technology. Moreover, in the presence of electricity price constraints, policy subsidies are necessary to incentivize the diffusion of CCS technology, with the suitable period for such incentives falling before 2040.

Using a Soccer team as a test bed for multi-agent systems simulation

The STMAS system is intended to imitate a soccer team and its behavior; we believe it can be used effectively as a test bed for multi-agent systems. It is constructed utilizing distributed agents that interact, communicate, and negotiate with each other to achieve the team objectives. It is based on the Jade simulation platform. The system is tested and compared to a pure soccer team using multiple MAS techniques. The results demonstrated that applying MAS techniques of negotiation and task distribution improves team performance, and STMAS is offered as an efficient test bed for new and distinct MAS techniques with varied scenario experiments. In addition, a mathematical model is created to compare the simulation results. Overall, STMAS provides a versatile and efficient MAS simulation and evaluation test bed. It is an excellent platform for comparing and evaluating various MAS approaches.

Multi-Objective Deep Reinforcement Learning for Crowd Route Guidance Optimization

In this study, we propose an improved version of Pareto deep Q-network (PDQN), a multi-objective deep reinforcement learning method, and attempt to demonstrate its effectiveness in a real-world problem such as crowd route guidance strategy optimization. Overcrowding during crowd movement can sometimes lead to accidents; therefore, it is imperative to guide crowds to move safely and efficiently. Safety and efficiency are conflicting objectives, and how to dynamically determine guidance can be formulated as a multi-objective sequential decision-making problem. PDQN is suitable for solving these problems, but its applicability to complex real-world problems has not been fully verified. To apply PDQN to real problems, we propose to adjust the parameters of PDQN and improve the action selection criteria during learning. A toy problem and a crowd guidance problem using multi-agent crowd simulation are adopted to evaluate the performance of improved PDQN. Experimental results show that the improved PDQN can search for good strategies compared with the original PDQN and can obtain better strategies than simplified strategies.

Can the introduction of CAVs mitigate social dilemmas causing traffic jams on highways?

This study explored whether the introduction of connected automated vehicles (CAVs) into human-driven vehicles (HVs) can mitigate the social dilemma caused by defective HVs that seek to overtake other vehicles by making frequent lane changes. In this regard, a cellular automata traffic model based on the revised S–NFS model, dealing with a mixed flow field composed of CAVs and HVs with multiple lanes, was established. We perused precise game theoretical analysis on the basis of the result obtained from traffic flow analysis through multi-agent simulation (MAS). As a result, the social efficiency deficit, quantifying the extent of the social dilemma behind such a specific traffic flow system, elucidates that the introduction of CAVs does not eliminate the social dilemma caused by defective HVs unless CAVs account for the vast majority of the mixed flow.

Neural-network-based parameter tuning for multi-agent simulation using deep reinforcement learning

This study proposes a new efficient parameter tuning method for multi-agent simulation (MAS) using deep reinforcement learning. MAS is currently a useful tool for social sciences, but is hard to realize realistic simulations due to its computational burden for parameter tuning. This study proposes efficient parameter tuning to address this issue using deep reinforcement learning methods. To improve compatibility with the tuning task, our proposed method employs actor-critic-based deep reinforcement learning, such as deep deterministic policy gradient (DDPG) and soft actor-critic (SAC). In addition to the customized version of DDPG and SAC for our task, we also propose three additional components to stabilize the learning: an action converter (DDPG only), a redundant full neural network actor, and a seed fixer. For experimental verification, we employ a parameter tuning task in an artificial financial market simulation, comparing our proposed model, its ablations, and the Bayesian estimation-based baseline. The results demonstrate that our model outperforms the baseline in terms of tuning performance, indicating that the additional components of the proposed method are essential. Moreover, the critic of our model works effectively as a surrogate model, that is, as an approximate function of the simulation, which allows the actor to tune the parameters appropriately. We have also found that the SAC-based method exhibits the best and fastest convergence, which we assume is achieved by the high exploration capability of SAC.