Agent-based Simulation Model Research Articles

Autonomous agent-based simulation modelling—A case study on a flexible GPU-card final assembly line

Market demands for high-tech products constantly evolve by product specification. To be competitive, a production system must be flexible and reconfigurable when facing mass customization. Flexible assembly line (FAL) enables mixed production with high efficiency. One example is graphic processing unit (GPU) cards. FAL requires comprehensive system design and scheduling to fully utilize the resources. This study adopts agent-based simulation (ABS) modelling for a FAL because of the abilities of ABS in flexibility and scalability. The proposed framework consists of three parts: real environment, virtual environment, and evaluation and analysis. This study uses agent-based simulation modelling to elaborate on sequencing and scheduling performances in the GPU-card assembly line. The Pareto frontier analysis is conducted to resolve conflicts between part tardiness and throughput.

An agent-based model for regional market penetration of electric vehicles in Brazil

Electrified vehicle (EV) are increasingly being adopted worldwide, making them promising elements in the decarbonisation process of the automotive sector. However, forecasting EV demand presents a significant challenge for policymakers worldwide. Our study proposes an agent-based simulation model that aggregates the behaviour of individual agents and generates cumulative adoption numbers through the Bass diffusion model. This approach enables using different parameters and extends previous research on actions to improve EV adoption rates in Brazil. Our investigation of the market introduction of EV using a multi-year dataset of EV buyers and marketed vehicle models reveals that acquisition cost is the primary driver of the adoption process. Still, barriers to electric mobility are also associated with vehicle-specific factors such as range, battery capacity, maintenance cost, recharging time, acquisition cost and number of EV sold. These findings emphasise the need to develop more effective policies and strategies for promoting EV adoption in Brazil, considering the factors that affect consumer behaviour and the adoption process. Doing so can accelerate the transition towards a more sustainable transportation system and reduce carbon emissions in the automotive sector.

Real-time multi-agent fleet management strategy for autonomous underground mines vehicles

ABSTRACT This paper proposes a real-time multi-agent fleet management strategy for autonomous underground mines vehicles. The fleet management strategy is based on multi-agent technology and includes a novel variation of the Contract-Net protocol. This paper also proposes a set of conflict management procedures to deal with the narrow nature of underground drifts, as well as the sequencing of trucks’ loading activities. This strategy is tested in a simulated environment based on an industrial case study. Both the strategy and the test environment were implanted using AnyLogic. More specifically, the fleet management activities addressed are dispatching, routing and traffic management of mining vehicles, which deal respectively with the assignment of the next destination to a vehicle that has just completed a task; the choice of the route to be followed to reach the selected destination; and traffic coordination in the underground transportation network, made up of one-lane bi-directional road segments. To evaluate the proposed solution, an agent-based simulation model of a Canadian underground gold mine is implemented with AnyLogic. Results show that the proposed coordination strategy outperform the one currently employed strategy by the mine under investigation.

Impact of school operating scenarios on COVID-19 transmission under vaccination in the U.S.: an agent-based simulation model

At the height of the COVID-19 pandemic, K-12 schools struggled to safely operate under the fast-changing pandemic situation. However, little is known about the impact of different school operating scenarios considering the ongoing efforts of vaccination. In this study, we deployed an agent-based simulation model to mimic disease transmission in a mid-sized community consisting of 10,000 households. A total of eight school operating scenarios were simulated, in decreasing order of restrictiveness regarding COVID-19 mitigation measures. When masks were worn at school, work, and community environments, increasing in-person education from 50% to 100% would result in only 1% increase in cumulative infections. When there were no masks nor contact tracing while schools were 100% in person, the cumulative infection increased by 86% compared to the scenario when both masking and contact tracing were in place. In the sensitivity analysis for vaccination efficacy, we found that higher vaccination efficacy was essential in reducing overall infections. Our findings showed that full in-person education was safe, especially when contact tracing, masking, and widespread vaccination were in place. If no masking nor contact tracing was practiced, the transmission would rose dramatically but eventually slow down due to herd immunity.

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.

Solving Problem with Generalists and Specialists: An Agent-based Simulation Model

Solving Problem with Generalists and Specialists: An Agent-based Simulation Model

An integrated agent-based simulation modeling framework for sustainable production of an Agrophotovoltaic system

The Agrophtovoltaic (APV) system is an alternative for sustainable crop production, where solar power is generated via Photovoltaic (PV) modules. Since both crop production and solar power generation activities are heavily dependent on dynamic environmental conditions, it is challenging to design an APV system based on accurate estimation of its performance. To this end, this study aims to introduce an agent-based simulation (ABS) framework integrated with polynomial regression and ridge regression. In particular, two agent types are devised, as follow: (1) The photovoltaic agent calculates electricity produced via PV modules, and estimates its profits; and (2) the crop production agent calculates crop harvests underneath the PV modules, and estimates their profits. To validate the proposed framework, field experiment data with five types of crops (i.e., corn, sesame, soybean, mung bean, and red bean) at the APV system at Jeollanam-do Agricultural Research and Extension Services in South Korea with three different shading ratios of 21.3%, 25.6%, and 32.0% have been used. In addition, for the sustainable operation of an APV system in terms of electricity generation as well as crop production, three climate change scenarios based on the shared socioeconomic pathway (SSP) are considered. The proposed framework identifies that the agrophotovoltaic system with 32% sharing ratio increases up to 20% of the total profit of normal farmland. As a result, the proposed framework enables the performance of an APV system under dynamic climate conditions to be accurately estimated, so that APV system designers can utilize it to identify a profitable long-term APV system.

An Agent-Based Simulation Model for Testing Novel Obesity Interventions in School Environment Design

Studying the impact of built-environment changes on human behavior and health is empirically challenging. To overcome this challenge, there has been increasing recognition of the usefulness of systems’ approaches for scenario testing to understand and prioritize such public health interventions. Limited research has shown a potential effect on the positive energy expenditure as a result of the use of dynamic furniture that promotes micromovement among school children. In this article, we aim to determine, via experiments in simulated populations of elementary school students, whether the use of dynamic furniture in the school environment has an impact on the childhood obesity over time. An agent-based model is developed to simulate three physical activity (PA) profiles over a period of six years with and without the dynamic furniture. Based on the prevalence trends from the experiments, there is no apparent impact of dynamic furniture use among boys, regardless of activity profile category. However, there are some pieces of evidence of differing trends among girls with a low PA profile. Although the intervention produces a marginal change in obesity prevalence in one population, the dynamic furniture is not a sole solution to the obesity epidemic. However, the use of dynamic furniture in schools may be a worthwhile component of prevention interventions.

A comprehensive study of agent-based airport terminal operations using surrogate modeling and simulation

Airport terminals are complex sociotechnical systems, in which humans interact with diverse technical systems. A natural way to represent them is through agent-based modeling. However, this method has two drawbacks: it entails a heavy computational burden and the emergent properties are often difficult to analyze. The purpose of our research is therefore to accurately abstract and explain the dynamics of airport terminal operations by means of computationally efficient and interpretable surrogate models, based on an existing detailed agent-based simulation model. We propose a methodology consisting of two stages. Stage I involves the development of faithful surrogates. A sample is collected according to an active learning strategy, upon which Gaussian process regression, higher-order polynomials, gradient boosting, and random forests are fitted. Stage II then applies state-of-the-art techniques from the emerging field of explainable artificial intelligence to interpret and understand these models. Both model-agnostic and model-specific methods are considered, and their results are synthesized in order to explain the emergent properties. We prove the efficacy of this approach by conducting two case studies on AATOM, an existing Agent-based Airport Terminal Operations Model. The first case study examines the total expenditure on discretionary activities, such as shopping and dining. A combination of poor staffing strategies and high occupancy rates on certain flights was found to disrupt the terminal journey of passengers on subsequent flights. As a result of these knock-on phenomena, less free time is left for discretionary activities, which has a negative effect on the total expenditure. The second case study examines the throughput of security checkpoints. While throughput increases with passenger numbers, a clear point was observed where the checkpoint reaches its maximum capacity. This leads to longer queues and therefore higher waiting times. It even goes so far as to put passengers at risk of missing their flight, especially with poor staffing strategies. Altogether, we clearly observed the preservation of emergent phenomena in surrogate models, and conclude that their combination with interpretable machine learning is an effective way to explain the dynamics of complex sociotechnical systems.

Optimisation of transport and technological system parameters of an agricultural enterprise in conditions of partial uncertainty

DOWNLOAD ARTICLE Optimisation of transport and technological system parameters of an agricultural enterprise in conditions of partial uncertainty Viacheslav Matsiuk, Viktoriia Opalko, Liliya Savchenko, Oleg Zagurskiy, Nadiia Matsiuk Received 07.04.2023, Revised 17.07.2023, Accepted 11.08.2023 Abstract At the stage of production of a wide range of agricultural products, to ensure the smooth operation of agricultural enterprises, it is necessary to solve the problems of fast and efficient delivery of relevant equipment, spare parts, and consumables with the rational use of available production resources. The research aims to improve the transport and technological system for the supply of orders in the form of consignments to meet the needs of the production activities of an agricultural enterprise. For this purpose, an agent-based simulation model was developed in the AnyLogic 8.7 environment using the Java compiler, since this toolkit allows simultaneously combining discrete-event and agent-based approaches. The model was implemented on the example of an enterprise of a separate subdivision of the National University of Life and Environmental Sciences of Ukraine “Agronomic Research Station”. As a result, a comprehensive, optimisation mathematical model of the supply of goods on an extensive network of road routes using the agricultural enterprise’s fleet of vehicles under conditions of partial uncertainty was obtained. In the course of experiments and calculations based on the real process of an agricultural enterprise, a range of values of the size of the truck fleet that meets the optimisation conditions has been found. It has been determined that the range of values close to the optimal size of the unloading truck fleet varies from 9 to 14 units. It has been established that the values of the optimisation criterion describing the average delivery time from the beginning of the need for an order to the moment of its delivery vary from 9.96 to 12.78 hours. The limit level of load of the transport and technological system is determined, at which the limit level of technological fault tolerance is ensured. It is 135 or more orders per year for each supplier with an estimated fleet of 12 trucks. The results of the study, such as the use of analytical tools and algorithms to optimise routes and allocate resources, can be used to improve transport efficiency, and help companies choose the most profitable and environmentally friendly routes for transportation

Biodiversity measures providing food and nesting habitat increase the number of bumblebee (Bombus terrestris) colonies in modelled agricultural landscapes

Individual biodiversity measures in agricultural landscapes such as flower strips are reported to promote the diversity and abundance of pollinating insects. Those biodiversity measures can be important food sources increasing the amount and continuity of pollen and nectar supply besides sometimes also offering nesting and hibernation sites. Still little is known about combined effects of biodiversity measures at the landscape scale and their long-term impact on pollinator population development. We used the agent-based simulation model BumbleBEEHAVE to investigate, if biodiversity measures have a positive effect on bumblebee populations at landscape scale in terms of the number of colonies per hectare. For this purpose, we chose Bombus terrestris, the most common bumblebee species in Germany, as target species. We used three real landscapes located in different regions of Germany as landscape settings for the simulations. The landscapes had strongly different farming systems regarding crop diversity and rotations and, consequently, different spatial structure and areal proportions of land-use types and semi-natural habitats. Between 2017 and 2020, distinct combinations of biodiversity measures, such as flower strips, flowering headland and fallows, were established by farmers on different areal proportions of the three landscapes. The biodiversity measures differed in plant-seed mixtures and, thus, the pollen and nectar supply by plants. We simulated the development of bumblebee colonies in the landscapes with and without the implemented biodiversity measures over eight years (four years in a twice repetition). We found that the implementation of biodiversity measures had a significant positive effect on the number of colonies. Further analysis showed that the pollen and nectar supplied by biodiversity measures had positive effects in all three landscapes, while the effect of additional nesting habitat differed among landscapes. Mass-flowering crops had little to no significant effect on the number of bumblebee colonies, whereas semi-natural habitats had a markedly positive effect. Our study underlines that not only biodiversity measures are likely to affect the bumblebee population, but that the overall landscape composition, particularly proportion of semi-natural habitats, is also important. So, to achieve high effectiveness of biodiversity measures, landscape context may be taken into account.

Mathematical Modeling of pedestrian flow through the obstacle bottleneck

For pedestrians in walking facilities, their movements are often obstructed by bottlenecks where the walkable widths are geometrically reduced. In previous research, to reproduce the influence of bottlenecks on pedestrian movements, agent-based simulation models have been widely applied. However, their high reliability on modeling rules and parameters requires calibration at the microscopic level, which makes them difficult to apply from an engineering perspective. Here, we applied a mathematical approach, which estimates the egress efficiency based on the density-flow rate fundamental diagram, to reproduce the egress time of pedestrians in our field experiments. Both the obstacle and the exit were considered bottlenecks in our experiments. It was indicated that with the same width, the obstacle bottleneck acted as an ‘ineffective’ bottleneck that did not affect the egress time when it was near the exit bottleneck, whereas acted as an ‘effective’ bottleneck when it was distant from the exit bottleneck. To reproduce this phenomenon, we applied a mathematical approach that abstracts the walking scenario as a scheme with the bottlenecks as links and different regions as nodes. As a result, the egress times under different layouts were reproduced successfully by introducing the density-flow rate fundamental diagram into the scheme. Furthermore, a reasonable range of the obstacle size and obstacle-exit width, under which condition the egress time is constant, was estimated. This study can be applied to estimate the egress time of the walking facilities by considering the fundamentals of pedestrian flows from a mathematical perspective, thus helping in the actual design of bottlenecks that could ensure efficient and safe pedestrian egress in walking facilities.

Landscape configurations of refuge areas that delay the evolution of resistance to Bt sugarcane: an agent based modeling approach.

Although transgenic crops expressing genes from the bacterium Bacillus thuringiensis (Bt) are considered to be an effective pest control method, reckless usage adds environmental pressure on a pest population to develop resistance to the protein over time. The use of small portions of non-Bt crop (refuge areas) limits the rate of resistance development. Strains of Bt sugarcane for the South African market are being developed, and a prerequisite to releasing such a product on the market is a recommendation on the size and layout of the refuge areas. In this article, an agent-based simulation model is used to test the effectiveness of different landscape configurations of refuge areas in Bt sugarcane against resistance development occurring in an associated lepidopteran pest population. Individual insects are modeled as agents on an underlying sugarcane field that can either be Bt or refugium. The model is applied to 2 hypothetical case studies, each focusing on a specific aspect of refugia planning. The first focuses on the size and distribution of refuge, and the second on the shape of the refuge. A conservative general recommendation of 30% per farm, planted in large blocks on farms, is made based on simulation results and what is currently known about the target pest species, to provide regulatory bodies as well as growers with a starting point on how to regulate and plan refuge areas in South African Bt sugarcane.

Ethnic segregation and spatial patterns of attitudes: studying the link using register data and social simulation

We theorize the causal link between ethnic residential segregation and polarization of ethnic attitudes within and between ethnic groups (e.g. attitudes towards immigration policies, multiculturalism, tolerance or trust in certain ethnic groups). We propose that the complex relationship between segregation and polarization might be explained by three assumptions: (1) ethnic membership moderates social influence–residents influence each other’s attitudes and their ethnic background moderates this influence; (2) spatial proximity between residents increases opportunities for influence; (3) the degree of ethnic segregation varies across space–and therefore, the mix of intra- and inter-ethnic influence also varies across space. We borrow and extend an (agent-based) simulation model of social influence to systematically explore how these three assumptions affect the polarization of ethnic attitudes within and between ethnic groups under the assumptions made in the model. We simulate neighborly interactions and social influence dynamics in the districts of Rotterdam, using empirically observed segregation patterns as input of our simulations. According to our model, polarization in ethnic attitudes is stronger in districts and parts of districts where mixing of ethnic groups allows for many opportunities to interact with both the ethnic ingroup and the outgroup. Our study provides a new theoretical perspective on polarization of ethnic attitudes by demonstrating that the segregation-polarization link can emerge as an unintended outcome from repeated intra- and inter-ethnic interactions in segregated spaces.

The Impact of Patient Infection Rate on Emergency Department Patient Flow: Hybrid Simulation Study in a Norwegian Case.

The COVID-19 pandemic put emergency departments all over the world under severe and unprecedented distress. Previous methods of evaluating patient flow impact, such as in-situ simulation, tabletop studies, etc., in a rapidly evolving pandemic are prohibitively impractical, time-consuming, costly, and inflexible. For instance, it is challenging to study the patient flow in the emergency department under different infection rates and get insights using in-situ simulation and tabletop studies. Despite circumventing many of these challenges, the simulation modeling approach and hybrid agent-based modeling stand underutilized. This study investigates the impact of increased patient infection rate on the emergency department patient flow by using a developed hybrid agent-based simulation model. This study reports findings on the patient infection rate in different emergency department patient flow configurations. This study's results quantify and demonstrate that an increase in patient infection rate will lead to an incremental deterioration of the patient flow metrics average length of stay and crowding within the emergency department, especially if the waiting functions are introduced. Along with other findings, it is concluded that waiting functions, including the waiting zone, make the single average length of stay an ineffective measure as it creates a multinomial distribution of several tendencies.

COVID-19: Agent-based simulation-optimization to vaccine center location vaccine allocation problem

This article presents an agent-based simulation-optimization modeling and algorithmic framework to determine the optimal vaccine center location and vaccine allocation strategies under budget constraints during an epidemic outbreak. Both simulation and optimization models incorporate population health dynamics, such as susceptible (S), vaccinated (V), infected (I) and recovered (R), while their integrated utilization focuses on the COVID-19 vaccine allocation challenges. We first formulate a dynamic location–allocation Mixed-Integer Programming (MIP) model, which determines the optimal vaccination center locations and vaccines allocated to vaccination centers, pharmacies, and health centers in a multi-period setting in each region over a geographical location. We then extend the agent-based epidemiological simulation model of COVID-19 (Covasim) by adding new vaccination compartments representing people who take the first vaccine shot and the first two shots. The Covasim involves complex disease transmission contact networks, including households, schools, and workplaces, and demographics, such as age-based disease transmission parameters. We combine the extended Covasim with the vaccination center location-allocation MIP model into one single simulation-optimization framework, which works iteratively forward and backward in time to determine the optimal vaccine allocation under varying disease dynamics. The agent-based simulation captures the inherent uncertainty in disease progression and forecasts the refined number of susceptible individuals and infections for the current time period to be used as an input into the optimization. We calibrate, validate, and test our simulation-optimization vaccine allocation model using the COVID-19 data and vaccine distribution case study in New Jersey. The resulting insights support ongoing mass vaccination efforts to mitigate the impact of the pandemic on public health, while the simulation-optimization algorithmic framework could be generalized for other epidemics.

Investigating crowd logistics platform operations for local food distribution

PurposeShort food supply chains have the potential to facilitate the transition to more sustainable food systems. Related distribution processes, however, can be challenging for smallholder and family farmers. To extend the market reach of farmers without the need for extensive investments, crowd logistics (CL) can be used. The purpose of this paper is to explore the benefits and trade-offs of implementing CL platforms in short food supply chains (SFSCs).Design/methodology/approachA decision support system (DSS) based on agent-based and discrete event simulation (DES) modelling is developed, which closely approximates the behaviour of customers and distribution processes at outlets. Different scenarios are explored to evaluate the potential of CL in rural and urban settings using the example of regions from Bavaria, Germany.FindingsResults show that CL can be used to increase the reach of farmers in SFSCs at the cost of minor food quality losses. Moreover, a difference between urban and rural settings is noted: An urban scenario requires less investment in the driver base, whereas the rural scenario shows a higher potential to increase market reach.Originality/valuePlatform-based food delivery services are still mostly unexplored in the context of SFSCs. This research shows that platform services such as CL can be used to support local agriculture and facilitate the distribution of perishable food items, introducing a simulation-based DSS and providing detailed results on various application settings; this research serves as a steppingstone to facilitate successful real-world implementations and encourage further research.

Agent-based societal impact modeling for infrastructure disruption and countermeasures analyses

Infrastructure disruptions can severely affect societal well-being. Quantifying the societal impacts of infrastructure disruptions and identifying appropriate countermeasures to mitigate them are critical for decision-makers. In this regard, this study measured the societal impact by considering people's achievement/intolerance of survival-related activities and proposed an agent-based societal impact simulation model (ABSISM) for quantitative estimation of the impact. The ABSISM incorporated dynamic interactions between households’ emergency behaviors, infrastructure disruption, and government countermeasures in typhoon disaster scenarios and designed decision rules for three types of agents: household, store, and shelter agents. Specifically, household agents acquire life-supporting resources (food and water) from store and/or shelter agents, and prioritize daily activities to perform under restricted resources by minimizing their experienced suffering level; store and shelter agents provide household agents with essential resources during a disaster. Additionally, the government's response measures (infrastructure recovery, forewarning information issuance, and distribution of supplies) were incorporated into the ABSISM as factors directly or indirectly influencing agents’ behaviors. Finally, this study presented an illustrative case study of Osaka to demonstrate the ABSISM's practical applicability and examines the effectiveness of government countermeasures on reducing the societal impact, which could provide insightful information for emergency management and promotion of sustainable cities.

A subcellular biochemical model for T6SS dynamics reveals winning competitive strategies.

The type VI secretion system (T6SS) is a broadly distributed interbacterial weapon that can be used to eliminate competing bacterial populations. Although unarmed target populations are typically used to study T6SS function in vitro, bacteria most likely encounter other T6SS-armed competitors in nature. However, the connection between subcellular details of the T6SS and the outcomes of such mutually lethal battles is not well understood. Here, we incorporate biological data derived from natural competitors of Vibrio fischeri light organ symbionts to build a biochemical model for T6SS at the single-cell level, which we then integrate into an agent-based model (ABM). Using the ABM, we isolate and experiment with strain-specific physiological differences between competitors in ways not possible with biological samples to identify winning strategies for T6SS-armed populations. Through in vitro experiments, we discover that strain-specific differences exist in T6SS activation speed. ABM simulations corroborate that faster activation is dominant in determining survival during competition. Once competitors are fully activated, the energy required for T6SS creates a tipping point where increased weapon building and firing becomes too costly to be advantageous. Through ABM simulations, we identify the threshold where this transition occurs in the T6SS parameter space. We also find that competitive outcomes depend on the geometry of the battlefield: unarmed target cells survive at the edges of a range expansion where unlimited territory can be claimed. Alternatively, competitions within a confined space, much like the light organ crypts where natural V. fischeri compete, result in the rapid elimination of the unarmed population.

Impact of self-learning based high-frequency traders on the stock market

In this paper we investigate the role of self-learning agents in multi-agent models of financial markets. We develop an agent-based simulation model of a financial market and, in addition to the agents with fixed strategies used in previous research, we introduce an agent with a self-learning strategy. To model the behavior of such an agent, we use deep reinforcement learning algorithms, namely deep deterministic policy gradient (DDPG). Next, we conduct a comparative analysis of the results of the constructed model with outcomes of previously proposed models, as well as with the characteristics of real market. To conduct comparative analysis, we use stylized facts of asset returns that allow us to evaluate and compare the characteristics of the markets. Our results show that a model with a self-learning agent gives a better approximation of the real market than a model with classic agents. In particular, unlike the model with classical agents, the model with a self-learning agent turns out to be not so heavy-tailed. Thus, we demonstrate that for a complete understanding of market processes simulation models should take into account self-learning agents that have a significant presence at modern financial markets.Finally, we present the python package11https://github.com/Kiri28/Agent-Based_Market_simulator., which was developed by us as part of the research implementation. This package allows to simulate the financial market, as well as create your own agents and evaluate their impact on the market.