It has often been shown that foresight information can improve fuel efficiency and comfort functions as well as active and passive safety systems. In this paper it will be shown how a database containing roadside information of a frequently driven route can be automatically generated and continually updated in the vehicle during the drive. The identified road characteristics such as curves, slopes, and speed information can be used as prediction or foresight information in subsequent drives along the route. The situation identification algorithms used in the system presented here base on standard sensors found in vehicles equipped with an electronic stability system and ABS. Additionally a positioning system such as GPS is required to define the geographical position of the identified road or traffic situation. An identified event such as a curve in the road or a steep uphill section can in terms of memory capacity be efficiently described with a set of attributes; geographical position, magnitude, number of observations, and date and time information of the observation. This information can be stored in a vehicle individual database with an analogous structure. During each drive this database can be continually extended and updated by comparing newly identified events with equivalent events already existing in the database. Especially for long distance transport, a sharing system of the learnt information would improve its functionality. As a result a new vehicle in a fleet or a new driver could immediately make use of already learnt roadside information collected from another driver/vehicle, without having to travel the route even once. Making use of further sensors, such as short and long range radar or cameras for lane detection and night-view, would open up the possibility for extended situation detection. In this case the information quality could be greatly improved. The development of the identification and comparison algorithms described here bases on real test drive data, performed through funding from the Geschwister-Heine Foundation and the Friedrich- und Elisabeth-Boysen Foundation.
An incremental learning method for foresight information used in predictive driving strategies
Eine Methode zum Lernen von Streckeninformationen für vorausschauende Fahrstrategien
2007
19 Seiten, 8 Bilder, 2 Tabellen, 22 Quellen
Conference paper
English
An incremental learning method for foresight information used in predictive driving strategies
Automotive engineering | 2007
|Automotive engineering | 2007
|Online Contents | 1996
Strategies for Gasoline Particulate Emission Control - A “Foresight Vehicle” Project
SAE Technical Papers | 2002
|