Behaviour Recognition for automated cars
|Title||Behaviour Recognition for automated cars|
|Summary||Behaviour Recognition for automated cars|
|Keywords||driver behavior, feature selection, prediction|
|Prerequisites||Artificial Intelligence, Machine Learning|
|Author||Cristofer Englund, Sławomir Nowaczyk, Anita Sant'Anna|
|Supervisor||Cristofer Englund, Sławomir Nowaczyk, Anita Sant'Anna|
There will be a smooth transition towards a fully automated traffic system that will continue for many years. This implies that there will be a mix of automated and manually driven vehicles in traffic. It is therefore important to understand the interactions between automated and manually driven vehicles. By initially studying the human behaviour and transferring that knowledge to automated vehicles we can create behaviours that are interpretable both for humans and for vehicles. This is fundamental knowledge in order to reduce the risk for accidents. The human should be able to understand the behaviour of an automated vehicle at least as good as he/she can interpret the behaviour of a manually driven vehicle.
At hand we have a 3D vision system that can capture and estimate vehicle speed, size, position and from this data motion paths are created and can be used for behaviour analysis. For this project the vision system should be mounted on the road infrastructure and measure vehicle behaviour in a road section where there are no formal rules but where the drivers apply social rules to interact with each other.
The thesis project will focus on the identification of features, or cues, that describe the intention of a driver, for example, in a narrow road section where vehicles can not meet (see figure below), what is the vehicle behaviour that leads to a negotiation of which vehicle passes first?
The project should include the following tasks
* Literature review * Data collection and preparation * Data analysis * Variable selection and modelling of driver (vehicle) behaviour * Writing report/paper