Simulating Complex Urban Behaviours With AI: Incorporating Improved Intelligent Agents in Urban Simulation Models

Solomou, Solon ORCID: https://orcid.org/0000-0003-1464-7836 and Sengupta, Ulysses ORCID: https://orcid.org/0000-0003-0342-3124 (2024) Simulating Complex Urban Behaviours With AI: Incorporating Improved Intelligent Agents in Urban Simulation Models. Urban Planning, 10. 8561. ISSN 2183-7635

Preview

Published Version Available under License Creative Commons Attribution. Download (1MB) | Preview

Abstract

Artificial intelligence is a transformational development across multiple research areas within urban planning. Urban simulation models have been an important part of urban planning for decades. Current advances in artificial intelligence have changed the scope of these models by enabling the incorporation of more complex agent behaviours in models aimed at understanding dweller behaviour within alternative future scenarios. The research presented in this article is situated in location choice modelling. It compares outcomes of two multi-agent systems, testing intelligent computer agent decision-making with selected behavioural patterns associated with human decision-making, given the same choices and scenarios. The majority of agent-based urban simulation models in use base the decision-making of agents on logic-based agent architecture and utility maximisation theory. This article explores the use of cognitive agent architecture as an alternative approach to endow agents with memory representation and experiential learning, thus enhancing their intelligence. The study evaluates the model’s suitability, strengths, and weaknesses, by comparing it against the results of a control model featuring commonly used logic-based architecture. The findings showcase the improved ability of cognitive-based intelligent agents to display dynamic market behaviours. The conclusion discusses the potential of utilising cognitive agent architectures and the ability of these models to investigate complex urban patterns incorporating unpredictability, uncertainty, non-linearity, adaptability, evolution, and emergence. The experiment demonstrates the possibility of modelling with more intelligent agents for future city planning and policy. Keywords: agent-based modelling; artificial intelligence; cognitive agents; complexity; household location choice; intelligent agents; market dynamics; planning tools; urban simulation