Level of Service estimation with in-vehicle sensor floating car data

Abstract

In the Smart In-Car project, in-vehicle signals obtained from the CAN (Controller Area Network) bus are collected along with the GPS position from 200 equipped vehicles. The main CAN bus signals in the available data set from the Smart In-Car project are: speedometer, brake usage, steering position, rpm, indicator usage and fuel consumption, but many more CAN bus signals exist, offering possibilities for many applications. In this thesis, the applicability of CAN bus signals for estimating the Level of Service (LOS) on motorways is assessed. The real CAN bus data of single equipped vehicles, obtained from the Smart In-Car project, is used to estimate the LOS, using loop detector data as ground truth. To assess the LOS estimation accuracy at higher penetration rates, simulation data is used. The available real CAN bus signals provide a correct LOS estimation of 37% with single vehicles. The simulation results show correct LOS estimation rates ranging from 52% at 1% penetration to 70% at 100% penetration. LOS estimation using only positional data (GPS) is found to outperform LOS estimation using CAN bus data for most penetration rates, showing that the CAN bus signals do not contribute to better LOS estimation. There are however many other applications which may benefit from CAN bus signals. These will require further research.