Multi-Agent Systems for Transportation Planning and Coordination

textabstractMany transportation problems are in fact coordination problems: problems that require communication, coordination and negotiation to be optimally solved. However, most software systems targeted at transportation have never approached it this way, and have instead concentrated on centralised optimisation. Multi-agent systems (MAS) are a different approach to building software systems. Such systems are assembled from autonomously interacting agents; agents are small software programs, which have some type of intelligence and individual behaviour. Communication and coordination (between agents) are the essential elements in the construction of MAS. The transportation domain is often referred to as a potential candidate for the application of MAS. In this dissertation, we discuss two MAS design cases related to the transport of containers. Both cases resulted in concrete prototypes, which let us evaluate a series of aspects important in applying MAS in transportation. We demonstrate the importance of a multi-method validation and evaluation approach. The prototypes were furthermore utilised as artefacts to discuss eventual implementation with future users and experts. One of our most important observations is that planning, as a function within supply chains, is about to go through a fundamental change. Like the mobile phone changed the way people coordinate in daily life, the concepts discussed in this dissertation have the potential to fundamentally change coordination in supply chains. As part of this fundamental change, a different perspective on certainty and uncertainty is essential.