Abstract

Intelligent transportation systems consisting of hundreds of individual vehicles which sense, decide, and act in the same-shared environment can be designed and controlled in two fundamentally different ways, using a centralized or distributed approach. The centralized approach implies an external system taking over the control of the vehicles and coordinating them, for instance by forming platoons, so that safety and fluidity are maintained. The distributed approach, instead, does not rely on any external control system and leaves the decisional autonomy to the individual vehicles. While the former approach has been shown to achieve a great degree of reliability, the latter represents an extremely appealing alternative because of its scalability and possible use in environments not endowed with external control systems. However, our human intuition and current engineering methods are not well adapted to design such distributed systems. In particular, when a certain group (or macroscopic) behavior is targeted, reverse engineering the individual (or microscopic) behavior is a non-trivial process. This process is even more complex and characterized by higher reliability requirements when each unit consists of a smart vehicle and a human being. This paper addresses design issues that could arise in the distributed approach and proposes an evolutionary computation methodology in order to help our intuition to devise the desired solutions. A simple case study concerning the placement and assessment of the number of proximity sensors is presented to support the discussion.