An Integrated Simulation Platform to Assist Policy-making for Autonomous Vehicles
Didac Busquets, Transport Systems Catapult, Christos Tsotskas, Transport Systems Catapult, Robin North, Transport Systems Catapult
We present a platform to address current requirements of urban traffic modelling to achieve interoperability between simulation components. We illustrate its use to assess the implications of autonomous vehicles in urban areas.
Current requirements for modelling urban mobility require the ability to integrate models and data from various sources and components. These may include microscopic simulators for vehicles or pedestrians, live feeds from real systems or activity based models, to name just a few. However, most software packages that simulate aspects of an urban environment work in a stand-alone, monolithic way, making it rather challenging to scale up and link to applications without changing the internals of participating components. In the few cases where components are coupled, they are usually bespoke interfaces that cannot be reused for other software packages.
Therefore, to ease the process of integrating diverse packages, we present the structural design of an open platform, so as to address strategic and operational requirements of urban traffic modelling. This builds on High Level Architecture (HLA) methods developed for military and aerospace applications. The main goal is to provide the means for academia and industry to integrate related systems in a transparent and time-synchronous way in accordance with well-established industrial standards. By using the developed platform, it is possible to achieve higher levels of integration and interoperability of existing and new urban models. This will enable traffic modellers to simulate more complicated real-world applications and simulate interactions that would not be achievable otherwise, while relying on existing, validated software packages.
An HLA platform relies on a publish-subscribe method to exchange data, where components use a "common dictionary" to represent the different entities that can be modelled (e.g. vehicles, pedestrians, roads) as well as the messages that can be sent from one component to another (e.g. start trip from one place to another, send information about congestion, etc.). Components can publish data through a common communication channel, without specifying who it is for. Similarly, components can subscribe to specific data types they are interested in, and they will receive such data independently of who published it. This independence is what makes the HLA platform really appropriate to achieve interoperability between components.
We are particularly interested in using this platform to combine urban micro-simulators with policy making. By integrating one or several components that take into account policies and regulations that affect the behaviour of the entities, we could easily simulate what effects a given policy would have on mobility. Such poilcies could be directly related to the use of the infrastructure (e.g. road closure, varying speed), or to the behaviour of some agents (e.g. pricing mechanism to promote the use of certain transport modes, opening times of shops). A system of this kind would be a very useful tool to assist decision makers to make more informed decisions and influence the future of urban traffic modelling in a continuous shared space.
In this paper we present the architecture of the platform and illustrate its use to assess the implications of autonomous vehicles on transport networks in urban areas. In particular, we have modelled a central area of Milton Keynes (UK) including its road network and shared space areas. The integration of several simulators into the platform allows us to analyse the interaction between conventional vehicles, driver-less vehicles and pedestrians, as well as to assess the effects of varying policies with respect to the use of shared spaces.
Association for European Transport