Multi-agent system is a rapidly developing field of distributed artificial intelligence that has gained significant importance because of its ability to solve complex real-world problems. It provides a highly flexible and modular structure, which incorporates the domain expertise in the system, to achieve the optimal solution. Multi-agent system also allows a problem to be divided into smaller sub-problems that require less domain expertise compared to solving the problem as a whole. In recent years, multi-agent system has gained significant attention in solving traffic signal control problems because of the advantages it offers in solving complex problems with uncertainties. In this paper, two different types of multi-agent architectures that have been implemented on a simulated complex urban traffic network in Singapore for adaptive intelligent signal control are discussed. The results obtained indicate the superior performance of the multi-agent signal controller in comparison to pre-timed and signal control methods which are currently in use.