Graph-based deep learning for representing events at the Large Hadron Collider

点击翻译按钮获取中文摘要

Connected and autonomous vehicles have the potential to make better use of the existing road infrastructure. This project will achieve this by suggesting personalised routes to users and optimising them based on personal preferences, and creating incentive mechanisms through the use of micropayments to follow the suggested routes. We envision a system where different users have different preferences, such as travel time, congestion and price, as well as environmental factors such as pollution. In addition, mechanism design will be used to incentivise users to follow the suggested route. In particular, the project will integrate the following strands of research:1. Congestion prediction. In order to suggest a certain route, prediction of congestion is needed. In order to do so accurately, the system will use routes and destinations shared by other users of the system to predict future congestion. At the same time, the system needs to consider privacy and ensuring a privacy-by-design approach. This part of the research will build on previous work published in [1].2. Suggestion optimisation. Based on the prediction and the personal preferences of individual users, the system will come up with route suggestions that maximise "social welfare" (i.e. the total utility of everyone). The novelty here compared to existing work is that, not only time, but other dimensions such as pollution levels and availability of EV charging stations, will be considered. Hence this is a multidimensional optimisation problem.3. Micro payments for incentives. This part of the project will use game theory/mechanism design techniques to create payments that incentivise users to adhere to the recommendation. This work builds on the mechanism design applications in related work, particularly on electric vehicle charging (see [2]). The novelty here is that the payments are "ex-post", i.e. they can change on the actual conditions observed. For example, if the user was rerouted to a "better" route but it turns out to be unexpectedly congested, this would change the payment.