D31.44 Decision System
SAFELANE
Confidential document
Executive Summary
The goal of the SAFELANE subproject is to develop a situation adaptive system for enhanced lane keeping support. The SAFELANE architecture consists of three main layers: perception, analyses and decision and action. The purpose of the described subtask is the development of the second layer called “decision system”. The development work started right after the end of the project specification and concept phases.
The decision system is structured in different modules.
Most likely path module predicts the future route of a vehicle in order to provide the correct map data. It uses a new algorithm based on pattern recognition and probability computation. This algorithm shows a good prediction performance.
The lane data fusion module incorporates different sources into the lane model estimation. The main lane sensor is the camera with the image processing. However, the data fusion algorithms includes map data, vehicle data and even radar objects trails to improve the availability and reliability of the lane model.
Static and dynamic approaches of trajectory prediction were compared to develop the trajectory estimation module. New variables are introduced to characterize a lane departure. The “lane predicted minimum distance” (LPMD) and “time to lane predicted minimum distance” (TLPMD) can be used as a replacement or an extension of a time to lane change (TLC) approach.
An open framework on the basis of a state machine concept is developed to describe a situation model, while adapting the lane keeping assistant to different situations. A concrete situation model is implemented for the SAFELANE demonstrator application. The model can be used for passenger cars as well as for commercial vehicles. Only the specific vehicle parameters must be adapted.
The decision model uses fuzzy functions to implement different driver support tasks. A decision tree is used to adapt the lane keeping support to different situations
|
