Large-scale Agent-based Models Research Articles

Large-scale agent-based simulation model of pedestrian traffic flows

Mobility patterns of pedestrians at a very high spatial and temporal resolution support urban planning strategies and facilitate a better understanding of the transport system. In this study, we develop an agent-based model that simulates disaggregated pedestrian traffic flows at a regional scale. This model is designed to overcome limitations of existing approaches in pedestrian traffic modelling by incorporating several decision processes that are defined by probabilistic rules. These rules include activity type, mode, and route choices. The results of a case study in Salzburg (Austria) show traffic flows concentrated in the city centre and along the river. Performed complexity analysis justifies the structure and logic of the modelled system by showing improvements in results during the stepwise implementation of model concepts. We also performed uncertainty analysis to provide information on the accuracy of the modelled results, showing strong and moderate correlations against observational data.

GEB v0.1: a large-scale agent-based socio-hydrological model – simulating 10 million individual farming households in a fully distributed hydrological model

Abstract. Humans play a large role in the hydrological system, e.g. by extracting large amounts of water for irrigation, often resulting in water stress and ecosystem degradation. By implementing large-scale adaptation measures, such as the construction of irrigation reservoirs, water stress and ecosystem degradation can be reduced. Yet we know that many decisions, such as the adoption of more effective irrigation techniques or changing crop types, are made at the farm level by a heterogeneous farmer population. While these decisions are usually advantageous for an individual farmer or their community, aggregate effects of those decisions can have large effects downstream. Similarly, decisions made by other stakeholders, such as governments, often have basin-wide effects and affect each farmer differently. To fully comprehend how the human–natural water system evolves over time and space and to explore which interventions are suitable to reduce water stress, it is important to consider human behaviour and feedbacks to the hydrological system simultaneously at the local household and large basin scales. Therefore, we present the Geographical, Environmental, and Behavioural (GEB) model, a coupled agent-based hydrological model that simulates the behaviour and daily bidirectional interaction of more than 10 million individual farm households with the hydrological system on a personal laptop. Farmers exhibit autonomous heterogeneous behaviour based on their characteristics, assets, environment, management policies, and social network. Examples of behaviour are irrigation, generation of income from selling crops, and investment in adaptation measures. Meanwhile, reservoir operators manage the amount of water available for irrigation and river discharge. All actions can be taken at a daily time step and influence the hydrological system directly or indirectly. GEB is dynamically linked with the spatially distributed grid-based hydrological model CWatM at 30′′ resolution (< 1 km at the Equator). Because many smallholder farm fields are much smaller than 1 × 1 km, CWatM was specifically adapted to implement dynamically sized hydrological response units (HRUs) at the farm level, providing each agent with an independently operated hydrological environment. While the model could be applied anywhere globally at both large and small scales, we explore its implementation in the heavily managed Krishna basin in India, which encompasses ∼ 8 % of India's land area and ∼ 12.1 million farmers.

Large scale simulation of pedestrian seismic evacuation including panic behavior

Understanding human behavior and predicting evacuation processes subsequently to an earthquake are critical aspects to provide valuable information to support disaster response activities. One possible and relevant aspect that characterizes human behavior in emergency situations is panic reactions, which can lead to dysfunctional and irrational responses by fugitives. Given the relevance of this component, this paper implements a panic behavior model in a large-scale agent-based model considering phenomenal parameters, such as seismic damage to the built environment, disruption of roads by earthquake-induced falling debris, and injuries of individuals. The proposed model is applied and tested in IdealCity, a virtual city simulation environment which resembles the geometry of the city of Turin, with about 900,000 inhabitants. The numerical simulations show that the inclusion of a panic behavior model increases the evacuation time despite the increased speed of agents during the evacuation process and this is probably caused by the fact that pedestrians tend to perform random actions before reaching their destination. Furthermore, existing human relationships between agents tend to convert the crowd from individual agents to group of agents that move together, and this behavior might affect the shelter saturation time of both shelters and hospitals as shown in the simulations analyzed in this paper.

Equity Analysis of Freight Transportation Using a Large-Scale Agent-Based Modeling Framework

Freight transportation is one of the largest contributors to the transportation sector's greenhouse gases (GHG) and noise pollution. The adverse effects of the industry are not distributed equally due to large spatial disparities, with minority and low-income populations more prone to live near freight-related facilities and truck routes. However, equity analysis focusing on freight transport is limited since empirical analysis requires costly data collection processes, and analysis in large-scale areas is challenging. The principal objective of this study is to conduct an equity analysis of the impact of freight-related trips in the Atlanta-Chattanooga-Knoxville region using a large-scale agent-based simulation framework. The main results show that low-income (less than $52k) and more heavily non-white population neighbourhoods are disadvantaged neighbourhoods in terms of noise and GHG emissions. The results and methods presented in this paper can be used to provide tools to develop transportation equity and social justice analysis in large networks and can be used to develop policies and regulations to reduce the impact of freight transportation on disadvantaged populations.

Simulation of rates of traffic accidents involving unmanned ground vehicles within a transportation system for the ‘smart city’

This article presents an approach to simulation of the rates of road traffic accidents involving unmanned ground vehicles within a multi-agent intelligent transportation system for the ‘smart city.’ A new simulation model of an intelligent transport system has been developed which makes it possible to significantly reduce the number of potential road traffic accidents (TAs) and implements the concept of socially-centered management of the urban economy. The software implementation of such a large-scale agent-based model was carried out using the FLAME GPU framework, which allows us to effectively parallelize the agents’ behaviour logic and consider their individual decision-making systems when modelling the spacial dynamics of an ensemble of unmanned ground vehicles (UGVs) interacting with other road users: the usual manned ground vehicles (MGVs), unexpected obstacles (e.g., pedestrians, etc.). Various scenarios of such agents’ behaviour in an intelligent transportation system are studied, including the occurrence of an accident under certain conditions (e.g., under the high speed and traffic intensity of UGVs, etc.) and various configurations of the digital road network (DRN). We determine the parameter values that provide for the individual decision-making system of UGVs remaining stable with respect to the characteristics of the external environment (including in extreme situations), ensuring the safety of other road users on the scale of the ‘smart city.’

DISTRIBUTED AGENT-BASED SIMULATION WITH REPAST4PY.

The increasing availability of high-performance computing (HPC) has accelerated the potential for applying computational simulation to capture ever more granular features of large, complex systems. This tutorial presents Repast4Py, the newest member of the Repast Suite of agent-based modeling toolkits. Repast4Py is a Python agent-based modeling framework that provides the ability to build large, MPI-distributed agent-based models (ABM) that span multiple processing cores. Simplifying the process of constructing large-scale ABMs, Repast4Py is designed to provide an easier on-ramp for researchers from diverse scientific communities to apply distributed ABM methods. We will present key Repast4Py components and how they are combined to create distributed simulations of different types, building on three example models that implement seven common distributed ABM use cases. We seek to illustrate the relationship between model structure and performance considerations, providing guidance on how to leverage Repast4Py features to develop well designed and performant distributed ABMs.

Gene drive mosquitoes can aid malaria elimination by retarding Plasmodium sporogonic development.

Gene drives hold promise for the genetic control of malaria vectors. The development of vector population modification strategies hinges on the availability of effector mechanisms impeding parasite development in transgenic mosquitoes. We augmented a midgut gene of the malaria mosquito Anopheles gambiae to secrete two exogenous antimicrobial peptides, magainin 2 and melittin. This small genetic modification, capable of efficient nonautonomous gene drive, hampers oocyst development in both Plasmodium falciparum and Plasmodium berghei. It delays the release of infectious sporozoites, while it simultaneously reduces the life span of homozygous female transgenic mosquitoes. Modeling the spread of this modification using a large-scale agent-based model of malaria epidemiology reveals that it can break the cycle of disease transmission across a range of transmission intensities.

Including Right-of-Way in a Joint Large-Scale Agent-Based Dynamic Traffic Assignment Model for Cars and Bicycles

Including Right-of-Way in a Joint Large-Scale Agent-Based Dynamic Traffic Assignment Model for Cars and Bicycles

PaCAR: COVID-19 Pandemic Control Decision Making via Large-Scale Agent-Based Modeling and Deep Reinforcement Learning.

Policy makers are facing more complicated challenges to balance saving lives and economic development in the post-vaccination era during a pandemic. Epidemic simulation models and pandemic control methods are designed to tackle this problem. However, most of the existing approaches cannot be applied to real-world cases due to the lack of adaptability to new scenarios and micro representational ability (especially for system dynamics models), the huge computation demand, and the inefficient use of historical information. We propose a novel Pandemic Control decision making framework via large-scale Agent-based modeling and deep Reinforcement learning (PaCAR) to search optimal control policies that can simultaneously minimize the spread of infection and the government restrictions. In the framework, we develop a new large-scale agent-based simulator with vaccine settings implemented to be calibrated and serve as a realistic environment for a city or a state. We also design a novel reinforcement learning architecture applicable to the pandemic control problem, with a reward carefully designed by the net monetary benefit framework and a sequence learning network to extract information from the sequential epidemiological observations, such as number of cases, vaccination, and so forth. Our approach outperforms the baselines designed by experts or adopted by real-world governments and is flexible in dealing with different variants, such as Alpha and Delta in COVID-19. PaCAR succeeds in controlling the pandemic with the lowest economic costs and relatively short epidemic duration and few cases. We further conduct extensive experiments to analyze the reasoning behind the resulting policy sequence and try to conclude this as an informative reference for policy makers in the post-vaccination era of COVID-19 and beyond. The modeling of economic costs, which are directly estimated by the level of government restrictions, is rather simple. This article mainly focuses on several specific control methods and single-wave pandemic control. The proposed framework PaCAR can offer adaptive pandemic control recommendations on different variants and population sizes. Intelligent pandemic control empowered by artificial intelligence may help us make it through the current COVID-19 and other possible pandemics in the future with less cost both of lives and economy. We introduce a new efficient, large-scale agent-based epidemic simulator in our framework PaCAR, which can be applied to train reinforcement learning networks in a real-world scenario with a population of more than 10,000,000.We develop a novel learning mechanism in PaCAR, which augments reinforcement learning with sequence learning, to learn the tradeoff policy decision of saving lives and economic development in the post-vaccination era.We demonstrate that the policy learned by PaCAR outperforms different benchmark policies under various reality conditions during COVID-19.We analyze the resulting policy given by PaCAR, and the lessons may shed light on better pandemic preparedness plans in the future.

Analysis and Visualization of High-Dimensional Dynamical Systems’ Phase Space Using a Network-Based Approach

The concept of attractors is considered critical in the study of dynamical systems as they represent the set of states that a system gravitates toward. However, it is generally difficult to analyze attractors in complex systems due to multiple reasons including chaos, high-dimensionality, and stochasticity. This paper explores a novel approach to analyzing attractors in complex systems by utilizing networks to represent phase spaces. We accomplish this by discretizing phase space and defining node associations with attractors by finding sink strongly connected components (SSCCs) within these networks. Moreover, the network representation of phase space facilitates the use of well-established techniques of network analysis to study the phase space of a complex system. We show the latter by introducing a new node-based metric called attractivity which can be used in conjunction with the SSCC as they are highly correlated. We demonstrate the proposed method by applying it to several chaotic dynamical systems and a large-scale agent-based social simulation model.

Simulation of migration and demographic processes using FLAME GPU

This article presents an approach to modeling migration and demographic processes using a framework designed for large-scale agent-based modeling – FLAME GPU. This approach is based on the previously developed simulation model of interaction between two communities: migrants and natives that is implemented in the AnyLogic simulation software. The model has had a low dimensionality of the discrete space representing the operating environment of the agent populations and a deterministic decision-making system of each agent. At the same time, the presence of multiple interactions between agents and transitions between their states determines a high computational complexity of such a model. The use of FLAME GPU makes it possible to conduct extensive simulation experiments with the model, mainly due to the parallelization of computational processes at the level of each agent, as well as the implementation of the mechanism of multiple computations using Monte Carlo techniques. The developed framework is used to study the impact of the most important parameters of the model (e.g., rate of migration, governmental expenditures on integration, frequency of creation of new workplaces, etc.) on the key outputs of the modeled socio-economic system (in particular, population size, share of migrants, number of assimilated migrants, GDP growth rate, etc.). The proposed approach can be used to develop decision-making systems for planning the hiring of new employees based on the forecast dynamics of migration and demographic processes.

Building and experimenting with an agent-based model to study the population-level impact of CommunityRx, a clinic-based community resource referral intervention.

CommunityRx (CRx), an information technology intervention, provides patients with a personalized list of healthful community resources (HealtheRx). In repeated clinical studies, nearly half of those who received clinical "doses" of the HealtheRx shared their information with others ("social doses"). Clinical trial design cannot fully capture the impact of information diffusion, which can act as a force multiplier for the intervention. Furthermore, experimentation is needed to understand how intervention delivery can optimize social spread under varying circumstances. To study information diffusion from CRx under varying conditions, we built an agent-based model (ABM). This study describes the model building process and illustrates how an ABM provides insight about information diffusion through in silico experimentation. To build the ABM, we constructed a synthetic population ("agents") using publicly-available data sources. Using clinical trial data, we developed empirically-informed processes simulating agent activities, resource knowledge evolution and information sharing. Using RepastHPC and chiSIM software, we replicated the intervention in silico, simulated information diffusion processes, and generated emergent information diffusion networks. The CRx ABM was calibrated using empirical data to replicate the CRx intervention in silico. We used the ABM to quantify information spread via social versus clinical dosing then conducted information diffusion experiments, comparing the social dosing effect of the intervention when delivered by physicians, nurses or clinical clerks. The synthetic population (N = 802,191) exhibited diverse behavioral characteristics, including activity and knowledge evolution patterns. In silico delivery of the intervention was replicated with high fidelity. Large-scale information diffusion networks emerged among agents exchanging resource information. Varying the propensity for information exchange resulted in networks with different topological characteristics. Community resource information spread via social dosing was nearly 4 fold that from clinical dosing alone and did not vary by delivery mode. This study, using CRx as an example, demonstrates the process of building and experimenting with an ABM to study information diffusion from, and the population-level impact of, a clinical information-based intervention. While the focus of the CRx ABM is to recreate the CRx intervention in silico, the general process of model building, and computational experimentation presented is generalizable to other large-scale ABMs of information diffusion.

How will mass-vaccination change COVID-19 lockdown requirements in Australia?

BackgroundTo prevent future outbreaks of COVID-19, Australia is pursuing a mass-vaccination approach in which a targeted group of the population comprising healthcare workers, aged-care residents and other individuals at increased risk of exposure will receive a highly effective priority vaccine. The rest of the population will instead have access to a less effective vaccine.MethodsWe apply a large-scale agent-based model of COVID-19 in Australia to investigate the possible implications of this hybrid approach to mass-vaccination. The model is calibrated to recent epidemiological and demographic data available in Australia, and accounts for several components of vaccine efficacy.FindingsWithin a feasible range of vaccine efficacy values, our model supports the assertion that complete herd immunity due to vaccination is not likely in the Australian context. For realistic scenarios in which herd immunity is not achieved, we simulate the effects of mass-vaccination on epidemic growth rate, and investigate the requirements of lockdown measures applied to curb subsequent outbreaks. In our simulations, Australia’s vaccination strategy can feasibly reduce required lockdown intensity and initial epidemic growth rate by 43% and 52%, respectively. The severity of epidemics, as measured by the peak number of daily new cases, decreases by up to two orders of magnitude under plausible mass-vaccination and lockdown strategies.InterpretationThe study presents a strong argument for a large-scale vaccination campaign in Australia, which would substantially reduce both the intensity of future outbreaks and the stringency of non-pharmaceutical interventions required for their suppression.FundingAustralian Research Council; National Health and Medical Research Council.

Modelling Singapore COVID-19 pandemic with a SEIR multiplex network model

In this paper, we have implemented a large-scale agent-based model to study the outbreak of coronavirus infectious diseases (COVID-19) in Singapore, taking into account complex human interaction pattern. In particular, the concept of multiplex network is utilized to differentiate between social interactions that happen in households and workplaces. In addition, weak interactions among crowds, transient interactions within social gatherings, and dense human contact between foreign workers in dormitories are also taken into consideration. Such a categorization in terms of a multiplex of social network connections together with the Susceptible-Exposed-Infectious-Removed (SEIR) epidemic model have enabled a more precise study of the feasibility and efficacy of control measures such as social distancing, work from home, and lockdown, at different moments and stages of the pandemics. Using this model, we study an epidemic outbreak that occurs within densely populated residential areas in Singapore. Our simulations show that residents in densely populated areas could be infected easily, even though they constitute a very small fraction of the whole population. Once infection begins in these areas, disease spreading is uncontrollable if appropriate control measures are not implemented.

Explaining herding and volatility in the cyclical price dynamics of urban housing markets using a large-scale agent-based model

Urban housing markets, along with markets of other assets, universally exhibit periods of strong price increases followed by sharp corrections. The mechanisms generating such non-linearities are not yet well understood. We develop an agent-based model populated by a large number of heterogeneous households. The agents’ behavioral rules are consistent with the concept of bounded rationality. The model is calibrated using several large and distributed datasets of the Greater Sydney region (demographic, economic and financial) across three specific and diverse periods since 2006. The model is not only capable of explaining price dynamics during these periods, but also reproduces the novel behavior actually observed immediately prior to the market peak in 2017, namely a sharp increase in the variability of prices. This novel behavior is related to a combination of trend-following aptitude of the household agents (herding) and their collective propensity to borrow. Trend-following behavior is found to be essential in replicating market dynamics.

A conceptual framework for modeling the supply side of mobility services within large-scale agent-based travel demand models

ABSTRACT The state-of-the-art of mobility services among large-scale agent-based travel demand models has focused predominantly on the demand side, whereas complexities of the supply side have been given much less attention. Conventional definitions must be revisited to establish solid foundations for model implementations. First, several mobility services are characterized from abstract attributes, by all agents involved, and by operational tasks involved in service provision. Then, a high-level generic service provision process is developed to allow instantiation of various services. Extensions for conventional travel demand model systems are also developed, including: a mobility services component containing service provider agentsand their operational tasks integrated logically into service provision processes; a fleet component containing vehicle agents of each service and a driver activity model when human-driven; a MaaS component containing mobility provider agents that coordinate and integrate mobility services. Generic modeling issues are also discussed, including treatment of time, data structures, and computational efficiency.

Application of bug navigation algorithms for large-scale agent-based evacuation modeling to support decision making

The global growth of urbanization has caused an increase of wildfires that devastate communities at wildland-urban interfaces. Agent-based evacuation simulation can be a powerful tool for emergency decision-makers to estimate evacuation times during rapidly propagating wildfires. However, as these models increase in size and complexity for urban areas, computation costs rise significantly, making these models less suitable for emergency situations. Although advances in high-performance computing have partially addressed the long execution time associated with large-scale models, the overall computation costs are still high considering the required infrastructure and technical knowledge. Alternatively, we can simplify certain components of the models such that accuracy will not be compromised greatly while computation speed is increased significantly. The authors suggest using the bug navigation algorithms, popular in the field of robotics, for the navigation of pedestrians. In the interest of finding the best candidate bug algorithm, a performance evaluation framework is also introduced. To demonstrate applicability, the evacuation of the city of Iquique, Chile, is simulated using the proposed approach. The results show that the proposed model is successful in estimating the evacuee arrival times, while execution time is reduced by orders of magnitude without a need for powerful processing resources.

Phase Transitions in Spatial Connectivity during Influenza Pandemics.

We investigated phase transitions in spatial connectivity during influenza pandemics, relating epidemic thresholds to the formation of clusters defined in terms of average infection. We employed a large-scale agent-based model of influenza spread at a national level: the Australian Census-based Epidemic Model (AceMod). In using the AceMod simulation framework, which leverages the 2016 Australian census data and generates a surrogate population of ≈23.4 million agents, we analysed the spread of simulated epidemics across geographical regions defined according to the Australian Statistical Geography Standard. We considered adjacent geographic regions with above average prevalence to be connected, and the resultant spatial connectivity was then analysed at specific time points of the epidemic. Specifically, we focused on the times when the epidemic prevalence peaks, either nationally (first wave) or at a community level (second wave). Using the percolation theory, we quantified the connectivity and identified critical regimes corresponding to abrupt changes in patterns of the spatial distribution of infection. The analysis of criticality is confirmed by computing Fisher Information in a model-independent way. The results suggest that the post-critical phase is characterised by different spatial patterns of infection developed during the first or second waves (distinguishing urban and rural epidemic peaks).

Large-scale Agent-Based Epidemic Models and Their Technical Implementation on Supercomputers

Large-scale Agent-Based Epidemic Models and Their Technical Implementation on Supercomputers

The safe development paradox: An agent-based model for flood risk under climate change in the European Union

With increasing flood risk due to climate change and socioeconomic trends, governments are under pressure to continue implementing flood protection measures, such as dikes, to reduce flood risk. However, research suggests that a sole focus on government-funded flood protection leads to an adverse increase in exposure as people and economic activities tend to concentrate in protected areas. Moreover, governmental flood protection can reduce the incentive for autonomous adaptation by local households, which paradoxically results in more severe consequences if an extreme flood event occurs. This phenomenon is often referred to as the ‘safe development paradox’ or ‘levee effect’ and is generally not accounted for in existing flood risk models used to assess developments in future flood risk under climate change. In this study we assess the impact of extreme flood events for the European Union using a large-scale agent-based model (ABM). We quantify how the safe development paradox affects (1) population growth and the increase in exposed property values, (2) the reduction in investments to flood-proof buildings as public protection increases, and (3) the increase in potential damage should a flood occur. For this analysis, we apply an ABM that integrates the dynamic behaviour of governments and residents into a large-scale flood risk assessment framework, in which we include estimates of changing population growth. We find that the impact of extreme flood events increases considerably when governments provide high protection levels, especially in large metropolitan areas. Moreover, we demonstrate how policy that stimulates the flood-proofing of buildings can largely counteract the effects of the safe development paradox.