China is one of the largest emerging markets for autonomous vehicles and our Shenjiang Roads model is the obvious choice when needing to test vehicles in this country. Customers can test their vehicle systems in a variety of very different situations which are all connected – you can turn directly off a busy highway into one of Shanghai’s small back streets.
The model consists of dual carriageway, large single carriageway and small backstreet China roads, presenting many opportunities for a variety of traffic scenarios, including other vehicles, pedestrians, cyclists etc.
This digital model features large, complex junctions (up to 5 lanes) with many traffic lights and pedestrian crossings. The roads also include multiple markings at junctions. The dense back-street single lane roads are full with shops and services. This presents hazards in the form of obstacles, such as shopping carts, which must be avoided. In contrast to the junctions, these roads feature very few road markings making navigation particularly challenging.
The roads in this model include a lot of information for perception systems to process. The streets are packed full of various objects at the roadside as well as overhead e.g. utility pipes. The shadows cast over the road by such objects are also a key factor when training and testing vehicle perception systems.
The road networks replicated in this digital twin are incredibly complex. The model is supplied with matching road networks which can be used to define the traffic on top of the model.
Our Shenjiang Roads model includes large complex junctions, extensive cycle lanes, pedestrian walkways throughout the model and a variety of road signs (overhead and roadside). The backstreets present considerable visibility challenges (vegetation, buildings, walls) for vehicles/cyclists/pedestrians to drive/walk in front of the vehicle making it a fantastic way to challenge vehicle sensors and systems.
This model offers our customers the opportunity to test their vehicle systems in a variety of situations which are all connected. Thousands of different test scenarios can be created without the need to add in scene varieties and the first local customer has now been using this model for over three years.