|Funding for:||UK Students, EU Students|
|Placed On:||19th August 2019|
|Closes:||30th September 2019|
The aim of this project is to build an AI system able to identify optimal swapping and charging locations for electric vehicles (EV). Charging/swapping locations are essential infrastructure for EV fleets.
Customers require these locations to be readily accessible at the times when they are needed (taking account of traffic conditions), and to provide the service that they require: e.g. rapid charging, or a replacement vehicle to swap into. Planning this infrastructure should take into account many factors such as (customer) population density, usage patterns by time of day/day of week, traffic conditions, as well as site availability and cost.
Providing these locations represents a major capital investment and so affects the viability of EV fleets as they are rolled out into new areas of the country. During this project we will developed Artificial Intelligence algorithms (such as Reinforcement Learning based on Markov decision processes) to construct a detailed model of customers’ behaviour to evaluate potential sites objectively (by large-scale simulation) in combination with traditional infrastructure planning criteria.
Solving this infrastructure planning problem requires models for the statistical distribution and behaviour of potential customers (the “arrivals process”). The project will integrate a wide range of data types including digital mapping, population density, places of work/home/retail/leisure, and dynamic data (e.g. traffic flows). The model will be calibrated using data collected from Evezy’s customers. Evezy currently manages a fleet of over 50 vehicles (expanding to 200 in 2019), all equipped with telemetry and camera systems. By using this model to simulate operation of proposed locations, we will be able to evaluate suitable multi-location configurations that minimise expected journey times for customers, while being cost effective.
Our approach will be able to incorporate dynamic resource allocation (e.g. relocation of vehicles/advice to drivers on availability) that may be required to capture the full complexity of the problem. The datasets used to support this project will be fully anonymised and will not include any sensitive information. This project will leverage expertise in AI (specifically sequential decision making under uncertainty and computer vision) within WMG Data Science group.
Type / Role: