MobiliTwin: from predictive to prescriptive analytics

这个创新项目的重点是开发一个综合平台，以优化公共交通的路线和调度。该解决方案将提供一个数据驱动的人工智能平台，为一个地区设计和运行优化的端到端公交服务，由动态响应公交（DRT）公交车和传统的智能公交车服务组合提供服务。重要的是，这两种类型的公交服务的人工智能优化将提供一个整体的公交解决方案，可以为现代地区的人口提供服务：无论是在密集的城市还是区域农村地区，为所有人提供包容性和无障碍的交通。这将提高效率，改善决策，在城市内部和人口不密集地区之间的交通需求平衡不断变化的挑战性环境中加强服务提供。 利用数字孪生模型和强化学习，该解决方案将基于历史数据、道路交通和事件信息，使用强化学习（RL）生成乘客需求，以创建数字孪生模型。数字孪生模型将通过生成大规模数据和使用人工智能进行训练来密切模仿现实世界的需求。该项目还包括实施基于强化学习的最佳路由模型。使用需求预测数据和真实的运营数据的持续学习将推动模型的不断发展。此外，该项目的重点是通过模型轻量化实现快速响应性能和实时优化路由，最终开发出由这些创新技术驱动的数字孪生模型。共同开发的解决方案将建立在项目合作伙伴Ciel & Alchera现有的数字孪生，公交路线和移动平台之上。项目合作伙伴已经在生产先进的机器学习方面取得了商业成功，并已经向这项合作研发的目标客户销售解决方案，并为这项创新的商业化制定了明确的路线图。

This innovative project focuses on developing a comprehensive platform to optimize routing and scheduling in public transportation. The solution will provide a data-driven AI platform to design and run an optimised end-to-end bus service for a region, served by a combination of Dynamic Responsive Transit (DRT) buses and traditional timetabled bus services. Importantly, this AI-powered optimisation of these two types of bus service delivery will provide a holistic bus solution which can serve a modern region's population: both in dense city and regional rural areas, providing inclusive and accessible transit for all. This will improve efficiency and decision-making for enhanced service delivery in a challenging environment where the balance of transit needs within cities and across less-densely populated regions is constantly in flux. Leveraging a digital twin and reinforcement learning, the solution will generate passenger demand using Reinforcement Learning (RL) based on historical data, road traffic, and event information to create a digital twin. The digital twin will closely mimic real-world demand by generating large-scale data and employing AI for training. The project also includes the implementation of a reinforcement learning-based optimal routing model. Continuous learning using demand prediction data and real operational data will drive ongoing model evolution. Additionally, the project focuses on achieving fast response performance and real-time optimal routing through model lightweighting, ultimately culminating in the development of a digital twin powered by these innovative techniques. The co-developed solution will be built on top of existing digital twin, bus routing and mobility platforms by project partners Ciel & Alchera. The project partners have established commercial success in productising advanced Machine Learning and already sell solutions to the target customers of this collaborative R&D, with a clear roadmap to commercialisation of this innovation.