Wai Kin Victor Chan

Agent-based simulation tutorial - simulation of emergent behavior and differences between agent-based simulation and discrete-event simulation

This tutorial demonstrates the use of agent-based simulation (ABS) in modeling emergent behaviors. We first introduce key concepts of ABS by using two simple examples: the Game of Life and the Boids models. We illustrate agent-based modeling issues and simulation of emergent behaviors by using examples in social networks, auction-type markets, emergency evacuation, crowd behavior under normal situations, biology, material science, chemistry, and archaeology. Finally, we discuss the relationship between ABS and other simulation methodologies and outline some research challenges in ABS.

Interaction metric of emergent behaviors in agent-based simulation

Agent-based simulation (ABS) has been a popular tool in various science and engineering domains. Simulating emergent behavior is one main usage of ABS. This paper investigates the use of interaction statistics as a metric for detecting emergent behaviors from ABS. An emergent behavior arises if this interaction metric deviates from normality.

An analysis of emerging behaviors in large-scale queueing-based service systems using agent-based simulation

This paper considers a large-scale service system consisting of a number of service areas (cells). Each cell contains a queueing model that operates continuously and independently from the queueing models in other cells. Each cell changes its state between alive and dead based on certain rules that depend on the queueing status of its own queue, the neighboring queues, and the whole community, while satisfying a constraint on the number of live cells in a neighborhood. The objective is to examine emerging behaviors from the interactions of the cells under various rules. Chaotic, deterministic, and in-between emerging behaviors are presented.

Duopoly electricity markets with accurate and inaccurate market goals

Electricity markets are complex systems due to their deregulation and restructuring. We develop an agent-based simulation model for a stylized electricity pool market and simulate the market as a repeated game. An online hill climbing with adjustment algorithm is applied to generator agents to guide them to bid strategically to reach their expected market share. It is observed that accurate (or genial) expected market goals lead to collusive behavior of generator agents with an equilibrium where their total profit is maximized. On the other hand, it is also found that inaccurate (or malicious) market goals could result in price war with an equilibrium where their profits are minimized.

Mathematical programming representations for state-dependent queues

Discrete-event dynamic systems with feedback, where the behavior of the system depends on the system state, are difficult to model due to the uncertainties and dependencies of system performance on the system state. Service systems, in particular, tend to exhibit this behavior where servers may work faster (or slower) when facing an increasingly long line of impatient customers. A common example is a state-dependent queue where the service rate depends on the queue size, which can change during service. In this paper, we present a mathematical programming representation for the sample path dynamics of a state-dependent queue, and illustrate its application in sensitivity analysis.

Adaptive routing and guidance approach for transportation evacuation

We propose an adaptive routing and guidance approach called Adjacent Node Score (ANS). This approach is integrated with an agent-based simulation model and avoids several common assumptions made in conventional evacuation models. ANS does not assume altruistic travelers and considers traffic interaction, variable link travel times, and their dependencies. This makes it more realistic than network flow models and stochastic routing algorithms. ANS can generate effective and good solutions at a low computational cost. It only requires local network information for routing and guidance. ANS can be easily implemented in practice. We test the ANS method on two networks and compare it with other four network routing strategies including the user-equilibrium condition, myopic, aggressive, and a naïve strategy that is based on static network information. Experimental results show that ANS can disperse highly concentrated traffic flows and reduce network clearance time compared with other methods.

Sensitivity analysis of linear programming formulations for G/G/m queue

Linear programming representations for discrete-event simulation provide an alternative approach for analyzing discrete-event simulations. This paper presents several formulations for G/G/m queues and discusses the applications and limitations of these formulations. We derive the relationship between these formulations. We then demonstrate the applications of these formulations in sample-path gradient estimation.

Bidding behaviors in duopoly electricity markets with aspirant market share goals

The deregulation and restructuring of electricity markets have created a variety of challenging research problems. In addition, due to the complexity of electricity markets, most of these research problems are not amenable to analytical methods. Agent-based simulation is an approach for simulating and analyzing complex systems with interacting autonomous agents. In this paper, we use an agent-based approach to study the following emergent problem related to electricity market share and competition: what happens if a market participant tries to reach the following two (sometimes conflicting) goals simultaneously, (1) reaching an aspirant market share goal and (2) maximizing profit? More interestingly, what happens if two such participants are competing with each other? The developed agent-based model allows us to examine how the market share goal and profit maximization goal together influence the bidding behaviors of generation companies (i.e. agents) in a day-ahead electricity auction market. It also reveals that conservative market share goals often lead to a collusive behavior and profit maximization. However, if every participant has an aggressive market-share goal, a price war would result. On the other hand, if agents bear unequal market-share goals (e.g. one aggressive and one conservative), one agent will become more profitable than the other. As a result, if both agents want to maximize profit, they will both bid aggressively, resulting in a price war. Therefore, the agent-based model produces results that may explain some real-world pricing outcomes. In addition, to benchmark our agent-based model and to demonstrate the effect of the market-share goal, we develop an analytical model without the market-share goal and compare its results with those from the agent-based model.