ALADIN

 

Active Learning based Automated Data Processing for Energy-efficient Driving Functions

 

Vehicle automation can contribute to energy-efficient mobility through predictive driving strategies and predictive powertrain control. The quality of the driving functions depends strongly on the quality of the data sets used for training and validation, so that increasingly large amounts of real driving data are continuously collected and preprocessed for specific applications. Due to the extreme amount of data, a manual selection of the relevant data from the entire measurement data is becoming increasingly impractical. In addition to a high expenditure of time and money, increased need for test drives as well as longer development cycles, the potential of vehicle automation in terms of safety and energy efficiency is thus only partially exploited. Intelligent, automated data pre-processing is therefore gaining importance, especially due to the strong increase in data-driven algorithms.

Active learning as an approach to automated data curation has the potential to identify relevant driving data without human expert knowledge. The method is based on the continuous application of existing prediction or perception algorithms to the measured sensor data, where the predictions of the algorithms are permanently compared with the real measured data. In this way, driving scenarios where high deviations occur are identified as informative data.

Within the ALADIN project, Active Learning will be exemplarily applied in the context of traffic prediction during in-vehicle data acquisition. For this purpose, a test vehicle will be equipped with environmental sensors and a programmable data logger on which the prediction models as well as the algorithms for Active Learning will be implemented. For final evaluation, data sets selected both manually and by Active Learning will be used to train the prediction model and the performances of the resulting models will be compared.

 

Funding

MWIDE