As industrial research in automated driving is rapidly advancing, it is of paramount importance to analyze field data from extensive road tests. This thesis presents a research work done in L3Pilot, the first comprehensive test of automated driving functions (ADFs) on public roads in Europe. L3Pilot is now completing the test of ADFs in vehicles by 13 companies. The tested functions are mainly of Society of Automotive Engineers (SAE) automation level 3, some of level 4. The overall collaboration among several organizations led to the design and development of a toolchain aimed at processing and managing experimental data sharable among all the vehicle manufacturers to answer a set of 100+ research questions (RQs) about the evaluation of ADFs at various levels, from technical system functioning to overall impact assessment. The toolchain was designed to support a coherent, robust workflow based on Field opErational teSt supporT Action (FESTA), a well-established reference methodology for automotive piloting. Key challenges included ensuring methodological soundness and data validity while protecting the vehicle manufacturers’ intellectual property. Through this toolchain, the project set up what could become a reference architecture for managing research data in automated vehicle tests. In the first step of the workflow, the methodology partners captured the quantitative requirements of each RQ in terms of the relevant data needed from the tests. L3Pilot did not intend to share the original vehicular signal timeseries, both for confidentiality reasons and for the enormous amount of data that would have been shared. As the factual basis for quantitatively answering the RQs, a set of performance indicators (PIs) was defined. The source vehicular signals were translated from their proprietary format into the common data format (CDF), which was defined by L3Pilot to support efficient processing through multiple partners’ tools, and data quality checking. The subsequent vi performance indicator (PI) computation step consists in synthesizing the vehicular time series into statistical syntheses to be stored in the project-shared database, namely the Consolidated Database (CDB). Computation of the PIs is segmented based on experimental condition, road type and driving scenarios, as required to answer the RQs. The supported analysis concerns both objective data, from vehicular sensors, and subjective data from user (test drivers and passengers) questionnaires. The overall L3Pilot toolchain allowed setting up a data management process involving several partners (vehicle manufacturers, research institutions, suppliers, and developers), with different perspectives and requirements. The system was deployed and used by all the relevant partners in the pilot sites. The experience highlights the importance of the reference methodology to theoretically inform and coherently manage all the steps of the project and the need for effective and efficient tools, to support the everyday work of all the involved research teams, from vehicle manufacturers to data analysts.
Big Data Management Towards Impact Assessment of Level 3 Automated Driving Functions
OSMAN, NISRINE
2021-07-28
Abstract
As industrial research in automated driving is rapidly advancing, it is of paramount importance to analyze field data from extensive road tests. This thesis presents a research work done in L3Pilot, the first comprehensive test of automated driving functions (ADFs) on public roads in Europe. L3Pilot is now completing the test of ADFs in vehicles by 13 companies. The tested functions are mainly of Society of Automotive Engineers (SAE) automation level 3, some of level 4. The overall collaboration among several organizations led to the design and development of a toolchain aimed at processing and managing experimental data sharable among all the vehicle manufacturers to answer a set of 100+ research questions (RQs) about the evaluation of ADFs at various levels, from technical system functioning to overall impact assessment. The toolchain was designed to support a coherent, robust workflow based on Field opErational teSt supporT Action (FESTA), a well-established reference methodology for automotive piloting. Key challenges included ensuring methodological soundness and data validity while protecting the vehicle manufacturers’ intellectual property. Through this toolchain, the project set up what could become a reference architecture for managing research data in automated vehicle tests. In the first step of the workflow, the methodology partners captured the quantitative requirements of each RQ in terms of the relevant data needed from the tests. L3Pilot did not intend to share the original vehicular signal timeseries, both for confidentiality reasons and for the enormous amount of data that would have been shared. As the factual basis for quantitatively answering the RQs, a set of performance indicators (PIs) was defined. The source vehicular signals were translated from their proprietary format into the common data format (CDF), which was defined by L3Pilot to support efficient processing through multiple partners’ tools, and data quality checking. The subsequent vi performance indicator (PI) computation step consists in synthesizing the vehicular time series into statistical syntheses to be stored in the project-shared database, namely the Consolidated Database (CDB). Computation of the PIs is segmented based on experimental condition, road type and driving scenarios, as required to answer the RQs. The supported analysis concerns both objective data, from vehicular sensors, and subjective data from user (test drivers and passengers) questionnaires. The overall L3Pilot toolchain allowed setting up a data management process involving several partners (vehicle manufacturers, research institutions, suppliers, and developers), with different perspectives and requirements. The system was deployed and used by all the relevant partners in the pilot sites. The experience highlights the importance of the reference methodology to theoretically inform and coherently manage all the steps of the project and the need for effective and efficient tools, to support the everyday work of all the involved research teams, from vehicle manufacturers to data analysts.File | Dimensione | Formato | |
---|---|---|---|
phdunige_4456069.pdf
accesso aperto
Tipologia:
Tesi di dottorato
Dimensione
1.59 MB
Formato
Adobe PDF
|
1.59 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.