Price Optimisation In Networks for Transport with Signals using Artificial Intelligence (PointsAI)

该项目将利用进化人工智能来解决道路运输网络中优化价格和信号的问题。设置道路价格和信号控制设置的能力是交通管理工具，可以用来优化网络中的交通流量。然而，价格和信号的变化会影响人们的旅行决策，因为个人的动机是尽量减少自己的旅行成本（包括旅行时间和金钱成本）。这意味着如果有更便宜的路线，旅行者会选择它们。这使得优化控制变量（例如价格和信号）的问题变得非线性且难以解决。网络中交通流量的优化程度需要根据某些目标进行评估，例如，最小化排队或减少污染，并且可以使用目标函数来精确量化目标达到的程度。遗传算法是一种人工智能技术，它利用自然选择的过程来找到问题的解决方案。它基于适者生存的理念；最适合的个体（即最有希望解决问题的个体）是那些最有可能生存和繁殖的个体。这种繁殖包括以随机方式从父母那里选择“染色体”，因此后代是两种“父母”解决方案的混合。如果重复这个过程，随着时间的推移，通常会改进解决方案。特定解决方案的性能（即特定的价格和信号设置）将通过在道路交通网络模型中采用这些控制设置并调整流量（朝着成本较低的路线）来评估，直到找到用户均衡流（当没有更多的路线交换时发生）。基于这些用户均衡流的目标函数的值（例如，它可以是网络中所有队列长度的总和）给出了解决方案性能的指示。为此目的所采用的道路交通网络模型将是RBM交通解决方案有限公司目前正在开发的计算机模拟模型。该模型可以快速找到用户平衡流，这使得它非常适合在实现遗传算法优化方法时需要的大量连续模型运行。

The project will utilise evolutionary artificial intelligence to address the problem of optimising prices and signals in road transport networks. The ability to set road prices and signal control settings are traffic management tools which can be utilised to optimise transport flows in the network. However, making changes to prices and signals will impact people's travel decisions, since individuals are motivated to minimise their own travel cost (which includes travel time as well as monetary cost). This means that travellers will swap to less costly routes should they be available. This makes the problem of optimising the control variables (for example the prices and signals) nonlinear and difficult to solve.The degree to which the traffic flows in a network are optimised needs to be assessed with respect to certain objectives, such as for example minimising queues or reducing pollution, and the use of an objective function can be used to precisely quantify the degree to which the objectives are met.A genetic algorithm is an artificial intelligence technique which uses a process of natural selection to find solutions to a problem. It is based on the idea of survival of the fittest; the fittest individuals (that is, the most promising solutions to the problem) being those most likely to survive and reproduce. This reproduction involves selection of 'chromosomes' from the parents in a random way, so that the offspring is a mix of the two 'parent' solutions. When repeated, this process generally leads to improving solutions over time.The performance of a particular solution (i.e. the particular price and signal settings) will be evaluated by adopting these control settings within a road traffic network model and adjusting flows (towards less costly routes) until the user equilibrium flows are found (which occurs when there is no more route swapping). The value of the objective function (which for example could be the sum of all the lengths of the queues in the network) based on these user equilibrium flows gives an indication of the performance of the solution. The road traffic network model utilised for this purpose will be the computer simulation model which is currently under ongoing development by RBM Traffic Solutions Ltd. This model can quickly find the user equilibrium flows, which makes it ideal for the large succession of model runs which will be needed when implementing the genetic algorithm optimisation approach.