Hybrid method integrating agent-based modeling and heuristic tree search for scheduling of complex batch processes

Agent-based modeling and simulation have evolved as a powerful tool for modeling complex systems, offering insights into emergent behaviors and interactions among diverse agents. Recently, integrating large language models into agent-based modeling and simulation presents a promising avenue for enhancing simulation capabilities. This paper surveys the landscape of utilizing large language models in agent-based modeling and simulation, discussing their challenges and promising future directions. In this survey, since this is an interdisciplinary field, we first introduce the background of agent-based modeling and simulation and large language model-empowered agents. We then discuss the motivation for applying large language models to agent-based simulation and systematically analyze the challenges in environment perception, human alignment, action generation, and evaluation. Most importantly, we provide a comprehensive overview of the recent works of large language model-empowered agent-based modeling and simulation in multiple scenarios, which can be divided into four domains: cyber, physical, social, and hybrid, covering simulation of both real-world and virtual environments, and how these works address the above challenges. Finally, since this area is new and quickly evolving, we discuss the open problems and promising future directions. We summarize the representative papers along with their code repositories in https://github.com/tsinghua-fib-lab/LLM-Agent-Based-Modeling-and-Simulation.

Autonomous vehicles (AVs) may have a transformative influence on transportation systems. AV use is expected to reduce crash, transport costs, and land use. However, the net effects of the autonomous vehicle are unknown because there are no fully-autonomous vehicles in reality. The simulation systems are indispensable to be built up not only for cost but also for obtaining massive data. An agent-based simulation is capable of simulating the entities in a complex system, such as the AV system. Considering the potential of agent-based models to complement existing models and the difficulty to understand, to duplicate and to compare different agent-based models, this paper review existing agent-based modeling of autonomous vehicles with the help of ODD protocol, which has been applied in the review of agent-based land-use change models and agricultural policy evaluation. The significant variables that researchers have taken into consideration in sensitivity analysis and in the different scenarios are explored in the paper. Fleet size is the primary variable researchers considered, and we also discussed the probable causes of different AV replacement rates. We have assessed the quality of reviewed articles to gain more acceptance from the agent-based modeling community. Most of the reviewed papers follow the ODD protocol of agent decision making. However, in order to improve the level of modeling transparency, the executable and related data shall be available, at least the data source and the programming language/simulation platform should be provided. Several recommendations for further research are also presented.

Many types of financial time series, most notably market returns, have been found to exhibit long-range memory as well as dramatic day-to-day swings that cannot be adequately represented by light-tailed distributions such as the normal distribution. In particular, this means that for such time series, the usual variance parameter (i.e., the sum of covariances at all time lags) is not defined because the covariance function does not converge to zero fast enough as the time lag increases. Moreover in such time series, often the tails of the marginal density converge to zero so slowly that higher-order marginal moments such as skewness and kurtosis fail to exist. Therefore conventional methods for analyzing simulation-generated time series cannot generally be applied to high-fidelity simulations of financial markets.Building on earlier work in fractal geometry and fractal time series, Mandelbrot et al. proposed the multifractal model of asset returns (MMAR) as an alternative to the ARCH models for analyzing time series exhibiting volatility clustering, long-range dependence, and heavy-tailed returns (Mandelbrot et al. 1997). They defined the multifractal spectrum as the renormalized probability density function of the Holder exponents observed in the time series; and they used the multifractal spectrum to measure the ability of MMAR to match the statistical properties of real data. In 2002 Kantelhardt et al. formulated multifractal detrended fluctuation analysis (MF-DFA), an algorithm for extracting the multifractal spectrum from a time series (Kantelhardt et al. 2002).Many economists have recently adopted agent-based simulation for modeling financial markets. Fads based on new products, shocks related to world events, scandals involving company leaders, or outright criminal activity can drastically change the processes and relationships governing a market. Already there is evidence that agent-based models yield more accurate approximations to the true or observed behavior in financial markets. Agent-based models exhibit emergent behaviors that have been linked to non-Gaussian interaction metrics and singularities in the time series they generate (Chan 2011). Farmer et al. constructed a agent-based model that is capable of exhibiting many of the statistical properties observed in real financial data simply due to the rules governing a double auction order book (Farmer et al. 2005).During the 20th century, economists relied heavily on Brownian motion to construct theoretical models of finance. The advantages of using Brownian motion stem from its foundation in the Gaussian distribution; and with advances in stochastic calculus and differential equations, it became a staple in models of empirical finance. The well-known Black-Scholes model for options pricing is a prime example of this application. Multifractal analysis has the ability to illuminate the underlying difference between a monofractal like Brownian motion and a considerably more diverse construction like the multifractal binomial measure. Thus, combining multifractal analysis with an agent-based model has the potential to provide insight into the underlying processes and behaviors of a dynamic economic system that is constantly changing and evolving.To analyze market-return time series exhibiting volatility clustering, long-range dependence, or heavy-tailed marginals, we exploit multifractal analysis and agent-based simulation. We develop a robust, automated software tool for extracting the multifractal spectrum of a time series based on MF-DFA. Guidelines are given for setting MF-DFA's parameters in practice. The software is tested on simulated data with closed-form monofractal and multifractal spectra as well as on observed data, and the results are analyzed. We also present a prototype agent-based financial market model based on the zero-intelligence model of Farmer et al. and analyze its output using MF-DFA. The figure below compares the multifractal spectrum of the output from the prototype model with that of the multifractal binomial measure. The ultimate objective is to expand this model to study the effects of microlevel agent behaviors on the macrolevel time series output as analyzed by MF-DFA. Finally we explore the potential for validating agent-based models using MF-DFA and thus being able to tune these models to the multifractal spectrum of empirical data.

Agent-based models provide a flexible framework that is frequently used for modelling many biological systems, including cell migration, molecular dynamics, ecology and epidemiology. Analysis of the model dynamics can be challenging due to their inherent stochasticity and heavy computational requirements. Common approaches to the analysis of agent-based models include extensive Monte Carlo simulation of the model or the derivation of coarse-grained differential equation models to predict the expected or averaged output from the agent-based model. Both of these approaches have limitations, however, as extensive computation of complex agent-based models may be infeasible, and coarse-grained differential equation models can fail to accurately describe model dynamics in certain parameter regimes. We propose that methods from the equation learning field provide a promising, novel and unifying approach for agent-based model analysis. Equation learning is a recent field of research from data science that aims to infer differential equation models directly from data. We use this tutorial to review how methods from equation learning can be used to learn differential equation models from agent-based model simulations. We demonstrate that this framework is easy to use, requires few model simulations, and accurately predicts model dynamics in parameter regions where coarse-grained differential equation models fail to do so. We highlight these advantages through several case studies involving two agent-based models that are broadly applicable to biological phenomena: a birth-death-migration model commonly used to explore cell biology experiments and a susceptible-infected-recovered model of infectious disease spread.

This article presents an overview of the agent-based modeling and simulation approach and its recent developments in transport fields, with the purpose of discovering the advantages and gaps and encouraging more valuable investigations and applications of agent-based models. We clarify the agent-based model from agents, the background of development, and the basic structure applied in transport systems. Then, the agent-based transport modeling toolkits are discussed. The applications of agent-based models in transport systems are reviewed in three time scale models followed by an additional discussion of hybrid modeling approaches. The extensive modeling of the beliefs, desires, learning, and adaptability of individuals and the optimization problems using agent-based models are explored. Besides, we point out some limitations in terms of calibration and validation procedure, agents’ behavior modeling, and computing efficiency. In conclusion, some recommendations are given and suggest potential and insightful directions such as Big Data and Digital Twin for future research.

With the proliferation of Web services and the evolution towards the Semantic Web comes the opportunity to automate various Web services tasks. A number of approaches have been proposed to tackle the problem of automatic web service composition. Most of them are inspired by the researches in AI planning. As heuristics plays a key role in the promotion of planning efficiency, the selection of heuristic mechanism is very important. In this paper, we propose the conception of composition distance, and propose a heuristic bidirectional search algorithm based on composition distance. The most contribution of this paper is : (1)propose the conception of composition distance; (2)a heuristic bidirectional search algorithm. The algorithm we proposed is effective and efficiency when applied in web service composition engine.

A heuristic search algorithm is given that determines and resolves the Hamiltonian circuit problem in directed graphs. The heuristic information of each vertex is a set composed of its possible path length values from the starting vertex, which is obtained by the path length extension algorithm. A detailed analysis of algorithm examples shows that the heuristic algorithm can greatly reduce the number of processing nodes compared to the backtracking algorithm without information; meanwhile, the algorithm’s time complexity is related to the number of loops in the directed graph. If a directed graph G(V, E) only contains loops that contain the starting point, then the Hamiltonian problem of G can be determined in polynomial time. The algorithm is implemented using C++ and Python programming, and all of the test graphs are generated by the random graph algorithm. The experiments show that the heuristic algorithm can reduce the number of extended nodes by at least 10% in directed random graphs with no or only a few loops. This heuristic algorithm can save time in applications that perform tasks while extending nodes because it reduce the number of extended nodes.

Reproducible modeling and simulation research has been identified as one of the Modeling and Simulation (M&S) Grand Challenge activities. Recently, uncertainty quantification has seen a renewed emphasis. While methods for verification and validation (V&V) have been widely developed for discrete-event simulations, newer simulation approaches such as the agent-based, agent-directed, and multi-agent simulation approaches introduce new V&V challenges. The active elements in these newer approaches have greater heterogeneity, e.g., every agent can be unique, with complex attributes and behaviors. Those behaviors can result in actions based on interaction with other agents, the environment, and even the outcome of simulated or artificial intelligence. The simulation spaces are often less constrained, e.g., rather than a network of servers and queues, the space can be continuous 2D Euclidian space with multiple associated geographic information systems (GIS) layers influencing the behavior of the actors. Over the last decade, a multitude of techniques has been used in agent-based modeling and simulation (ABMS) to perform V&V as well as replication and reproducibility (R&R) of the models. In this chapter, we present an overview of contributions in V&V, quality assurance (QA), and R&R of simulation studies, with special focus on ABMS. We also discuss the lessons learnt in V&V and replication from a series of simulation experiments using agent-based models (ABMs).

Agent-based modeling and simulation is a simulation paradigm that allows focusing on individuals, their interactions, and the complex behavior that emerges from them. Agent-based simulations are typically developed in simulation platforms that provide features related to agents. One such platform is NetLogo, to which a reinforcement learning extension was made available recently. The extension provides commands for using the Q-Learning algorithm, but no evaluation on whether it simplifies the development of simulations is available. This paper presents a quantitative evaluation on using the extension in two simulations: the classic cliff walking problem; and a real-world, adaptive traffic signal control (ATSC) simulation. Results show that the size of simulations source code developed using the extension is smaller than those developed without using it, giving evidence that the extension simplifies the development of simulations

Agent-based modeling and simulation in architecture

The scheduling of batch process is a kind of NP-complete problem and has been the interest of research for many years. Various methods have been applied to solve such problems and simulated annealing (SA) is the most efficient algorithm. But SA is very slow when the problem size is large. In this work, particle swarm optimization (PSO) was applied to solve the scheduling problem of multiproduct batch process. The results show that PSO is a powerful method for solving the batch process scheduling problems and is superior to the widely used simulated annealing.

Landowner characteristics influence his/her willingness to change landuse practices to provide more or less environmental benefits. However, most studies of agricultural/environmental polices identify landowners as homogenous. And, the primary cause of failure of many environmental and other polices is the lack of knowledge on how humans may respond to polices based on changes in their behavior (Stern, 1993). From socioeconomic theory and empirical research, landowners can be identified as individuals who make agricultural landuse decisions independently based on their objectives. Identifying possible classes of landowners, assessing how each would potentially respond to policy alternatives, and the resulting pattern of land uses in a watershed or a riparian corridor would be very useful to policy makers as they evaluated alternatives. Agricultural landscapes are important producers of ecosystem services. The mix of ecosystem services and commodity outputs of an agricultural landscape depends on the spatial pattern of land uses emerging from individual land use decisions. However, many empirical studies show that the production of ecosystem services from agricultural landscapes is declining. This is consistent with research conducted over the last few decades showing there is a narrow range of social circumstances under which landowners are willing to make investments in the present to achieve public benefits in the future through investing in natural capital resulting in public goods which are frequently produced as ecosystem services. In this study an agent-based model within a watershed planning context is used to analyze the tradeoffs involved in producing a number of ecosystem services and agricultural commodities given price and policy scenarios while assuming three different types of agents in terms of their goals. The agents represent landowners who have been divided into a number of different groups based on their goals and the size of their farm operations. The multi-agent-based model is developed using a heuristic search and optimization technique called genetic algorithm (GA) (Holland), which belongs to a broader class of evolutionary algorithms. GAs exhibit three properties (1) they start with a population of solution, (2) they explore the solution space through recombination and mutation and (3) they evaluate individual solutions based on their appropriate fitness value(s), for example given profit maximizing agents this would be gross margin. A GA is a heuristic stochastic search and optimization method, which works by mimicking the evolutionary principles and chromosomal processing in natural genetics. The three economic agents that are modeled are based on variations in their objective functions and constraints. This study will help in identifying the tradeoffs associated with various agents in the provision of ecosystem services and agricultural commodities. The agent model developed here will help policy and decision maker identify the various agents within the watershed and assess various policy options based on that information. The study will also help to understand the interaction and feedback between the agents and their environment associated with various policy initiatives. The results of the study indicate that the agent model correctly predicts the actual landuse landcover map by 75 percent. Key words: Multifunctional agriculture, Agent based modeling, Genetic Algorithm.

Loading and unloading activities generate nearly fifty per cent of the total cost in port. The loading and unloading process of the container at the port is considered as a complex process since it involves several interrelated components, such as ships, cranes, and trucks. The uncertainty of these component activities might impact the loading and unloading time and cost. Agent-based modelling and simulation (ABMS) approach is a method for analyzing and modelling a complex system. This study aims to simulate the unloading process to determine a strategy to reduces unloading process time in the largest port in Indonesian using ABMS approach. The results show that the agent-based simulation approach is feasible to be applied in port activities. This approach can assist decision-makers in predicting the number of facilities that must be used to minimize processing time.

Prospective Health Technology Assessment allows early decision making for innovative health care technologies. The main idea is to combine available domain knowledge with advanced simulation techniques in order to predict the effects of medical products and to find bottlenecks and weaknesses within the health system. In our recent publications a hybrid simulation approach with System Dynamics and Agent-Based Modeling has been presented. Hospital workflows have been modeled by state charts within agent behavioral models and have to be instantiated each time an agent is entering a hospital. This paper presents a mechanism to generate agents dynamically from SD models and extends the previously presented hybrid approach by process-oriented Discrete Event Simulation for hospital modeling. It connects processes to health care institutions and not to persons traversing them. Two extended example case studies show potentials for medical decision making using the three simulation paradigms in a common environment.

Agent-based simulations are an increasingly popular means of exploring and understanding complex social systems. In order to be useful, these simulations must capture a range of aspects of the modeled situation, each possibly requiring distinct expertise. Moreover, different paradigms may be useful in modelling, ranging from those that use many lightweight reactive agents, to those that use cognitive agents, to those that focus on agent teams and organisational structures. There is need for an architecture which supports the development of a large simulation, through the integration of separately developed modules. This paper describes a framework and architecture which facilitates the integration of multiple agent-based simulations into a single global simulation. This architecture naturally supports distributed simulation and incremental development, which are ways of addressing the computational and conceptual complexity of such systems. In this paper we focus particularly on how to ensure proper management of simulation data that is affected by agents in different modules, at the same logical time. We also provide some preliminary performance evaluation addressing scalability, as well as a comparison of how other available systems handle the issue of shared data.