Car Parking as a Game Between Simple Heuristics

Except for reformatting and the addition of a reference list, the authors believe that this MS does not differ from that published as: Hutchinson, J. C., Fanselow, C. & Todd, P. M. (2012).

Abstract

Abstract Selecting a parking space is a sequential-search problem, with the pattern of spaces available to search determined by where others have parked. Earlier optimality models of parking ignored this game-theoretic aspect, unrealistically assuming random occurrences of spaces. This chapter instead simulates populations of cars: Agents (drivers) may accept any unoccupied space as they proceed down a dead-end street toward the destination; otherwise they take the first as they drive back out. Several simple decision heuristics are considered and Nash equilibria found with parameter values that are sensitive to conditions and performance criteria. An evolutionary algorithm is used to implement a competition between the heuristics. The winning heuristic exploits emergent environment structure: Because adjacent sites often fill sequentially, they empty at similar times, so spaces that drivers have recently encountered predict more spaces ahead. High car densities arising far from the destination also leads this heuristic to reject spaces until within a fixed distance of the destination.