An Evolutionary Approach to the Optimisation of Autonomous Pod Distribution for Application in an Urban Transportation Service

For autonomous vehicles (AVs), which when deployed in urban areas are called “pods”, to be used as part of a commercially viable low-cost urban transport system, they will need to operate efficiently. Among ways to achieve efficiency, is to minimise time vehicles are not serving users. To reduce the amount of wasted time, this paper presents a novel approach for distribution of AVs within an urban environment. Our approach uses evolutionary computation, in the form of a genetic algorithm (GA), which is applied to a simulation of an intelligent transportation service, operating in the city of Coventry, UK. The goal of the GA is to optimise distribution of pods, to reduce the amount of user waiting time. To test the algorithm, real-world transport data was obtained for Coventry, which in turn was processed to generate user demand patterns. Results from the study showed a 30% increase in the number of successful journeys completed in a 24 hours, compared to a random distribution. The implications of these findings could yield significant benefits for fleet management companies. These include increases in profits per day, a decrease in capital cost, and better energy efficiency. The algorithm could also be adapted to any service offering pick up and drop of points, including package delivery and transportation of goods.