WO2023247092A1 - Conception d'un système de commande de conduite automatique dans des simulations massivement parallèles - Google Patents

Conception d'un système de commande de conduite automatique dans des simulations massivement parallèles Download PDF

Info

Publication number
WO2023247092A1
WO2023247092A1 PCT/EP2023/060623 EP2023060623W WO2023247092A1 WO 2023247092 A1 WO2023247092 A1 WO 2023247092A1 EP 2023060623 W EP2023060623 W EP 2023060623W WO 2023247092 A1 WO2023247092 A1 WO 2023247092A1
Authority
WO
WIPO (PCT)
Prior art keywords
driving control
data
control system
road users
automatic driving
Prior art date
Application number
PCT/EP2023/060623
Other languages
German (de)
English (en)
Inventor
Nico Weber
Christoph THIEM
Ulrich Eberle
Original Assignee
Stellantis Auto Sas
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Stellantis Auto Sas filed Critical Stellantis Auto Sas
Publication of WO2023247092A1 publication Critical patent/WO2023247092A1/fr

Links

Classifications

    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0108—Measuring and analyzing of parameters relative to traffic conditions based on the source of data
    • G08G1/012—Measuring and analyzing of parameters relative to traffic conditions based on the source of data from other sources than vehicle or roadside beacons, e.g. mobile networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/15—Vehicle, aircraft or watercraft design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/008—Artificial life, i.e. computing arrangements simulating life based on physical entities controlled by simulated intelligence so as to replicate intelligent life forms, e.g. based on robots replicating pets or humans in their appearance or behaviour
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/09—Supervised learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/092—Reinforcement learning
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0125—Traffic data processing
    • G08G1/0129—Traffic data processing for creating historical data or processing based on historical data
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0125—Traffic data processing
    • G08G1/0133—Traffic data processing for classifying traffic situation
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/16—Anti-collision systems
    • G08G1/166—Anti-collision systems for active traffic, e.g. moving vehicles, pedestrians, bikes

Definitions

  • the invention relates to a method for designing and validating an automatic driving control system of a vehicle.
  • simulation-based development and testing of automated vehicles with an automatic driving control system in particular the validation of a driving control function of an automatic driving control system for the automatic operation of a vehicle, is only possible with the help of simulation if the respective simulation modules demonstrably produce realistic results.
  • This also applies, among other things, to modeling traffic dynamics, i.e. simulating the behavior of other road users.
  • a realistic representation of the environmental dynamics is therefore a key component for finding potentially critical scenarios for an automated vehicle.
  • the object of the invention is to improve the design and validation of an automatic driving control system of a vehicle.
  • a first aspect of the invention relates to a method for designing and validating a automatic driving control system of a vehicle, comprising the steps:
  • the data of the road users observed in reality can, on the one hand, be recorded with stationary sensors, especially in particularly interesting areas such as urban intersections, but on the other hand can also be recorded from an advantageous bird's eye view, for example with a balloon, a quadrocopter or a small-scale fixed-wing aircraft or another unmanned one aircraft can be detected.
  • This means that the real behavior of real road users is known and can form a realistic approach for the initial database for the simulations.
  • the respective associated scenario is recorded accordingly, for example in order to provide not only the real road users but also the background scenario such as intersection data for the simulations.
  • This scenario is advantageously included in the respective simulation in order to create a simulation that is as realistic as possible.
  • the respective associated scenario can also include weather data, for example prevailing light intensity, wind strength, precipitation, fog, etc.;
  • the trajectories of the real observing road users include a trajectory and associated time information, so that the temporally relative movement of the road users can be reproduced in the agent models of the simulations.
  • Preparing the data from real road users to ensure the completeness of the data ensures that the entirety of the traffic dynamics relevant to the automated vehicle can be observed within the identified traffic sections.
  • the calibration and validation of the simulation environment can then take place, which is able to model and simulate the dynamics of all relevant road users for the corresponding traffic area with sufficient validity.
  • the real road users are then modeled as traffic agents so that the actually observed behavior of the real road users can be depicted in the respective simulation and the respective simulation can depict dynamic scenarios.
  • the calibrated and validated simulation environment is the basis for the subsequent process step of massively parallel long-term simulation, in which the modeled traffic agents are used as the initial database with the addition of synthetic data from virtual road users.
  • long-term simulations are carried out with the initial database in multiple parallelization of the simulations and with a time scale factor greater than one for real time, integrating the automatic driving control system with simulation data instead of sensor data.
  • the fact that there is a time scale factor greater than one compared to real time means that the simulations are carried out significantly faster than real time and therefore huge amounts of data from very long simulated travel times can be processed in simulations in a relatively short time - hence the expression "long-term simulation” .
  • the multiple parallelization also contributes to this, meaning that several simulations are carried out in parallel at the same time.
  • a respective variation of the initial database or a respective set of variations is carried out with each individual simulation.
  • the validation data set is used to check whether the automatic driving control system operates safely across a very large number of realistic scenarios.
  • the generated validation data set defines the entire surrounding traffic dynamics relevant to the automated vehicle. Even if this methodology is particularly suitable for validating an algorithm of the driving control function of the automatic driving control system and thus for securing situational understanding, decision-making, regulation and actuation of the automatic driving control system, the use of the database to support the securing of the environment detection/perception of the automated vehicle is also possible. This can be done, for example, by combining the validation data set with a realistic visualization simulator.
  • the modeling of, for example, material properties and reflections of the relevant static and dynamic objects in the environment of the automated driving function is carried out by the visualization simulator, while the varied initial database defines the relevant traffic dynamics (type of movement of the dynamic objects) in the corresponding scenario.
  • the synthetic data of virtual road users are generated by completely synthetic generation without reference to the real road users.
  • the synthetic data of the virtual road users are generated partly by modifying the data of the real road users and partly by completely synthetic generation without reference to the real road users.
  • the variations of the initial database include a varying ratio between modeled human traffic agents and virtual traffic participants.
  • Varying the relationship between the virtual road users and the modeled human traffic agents allows a sensitivity analysis regarding the degree of penetration of automated road users, especially automated vehicles.
  • the free variation of this ratio is in a physically real situation not possible.
  • This embodiment can advantageously be used to determine the sensitivity of the automatic driving control system to the proportion of automatically guided vehicles in the traffic area around the vehicle. This is of great interest because automatically guided vehicles behave differently than human-driven vehicles.
  • the proportion of automatically driven vehicles in relation to all road users is also called the penetration level of automated vehicles.
  • the number of virtual road users in the output database is greater than the number of real road users used.
  • strategic and tactical, or strategic, tactical and operational simulation elements for the vehicle are carried out in the simulations carried out.
  • the hybrid, varied initial database is used in particular for training at the strategic, tactical or operational level.
  • An example of the strategic level is the sensible choice of route based on a defined start and destination.
  • An example of the tactical level is the execution of the maneuver prediction module in the form of the future trajectories of the road users surrounding the automated vehicle in a specific traffic situation.
  • the behavior/trajectories of the real road users when safely passing the respective scenario can be used as ground truth.
  • the operational level relates in particular to a respective vehicle management function.
  • Another obvious use case is the training of a reinforcement learning agent, which is exposed to the corresponding traffic dynamics and, depending on its reaction, is rewarded or punished for its behavior and can thus learn safe behavior across the relevant traffic scenarios.
  • a sensitivity analysis is determined via the effect of variations in the output database on the respective reaction of the automatic driving control system and the result of the sensitivity analysis is used for a targeted adaptation process of parameters of a driving control function of the automatic driving control system.
  • This sensitivity analysis not only looks at a single data point in the simulation space, but varies the output data base around a data point to get a sense of the impact of these changes in multiple directions on the response of the automatic driving control system. This applies in particular to the number of modeled automated road users in relation to the total number of other road users simulated in the simulation. This level of penetration is of particular interest in sensitivity analysis.
  • an environment detection/perception of the automated vehicle is also validated using this embodiment.
  • a sensor unit with a connected perception module is preferably considered, with the sensor unit being artificially fed data from the simulator instead of real data in order to validate the environment detection and perception of the automated vehicle.
  • Material properties and reflections of the relevant static and dynamic objects in the environment of the automated driving function can also advantageously be used by a visualization module in the simulation in order to generate realistic states of detectable sensor data.
  • optical material properties and/or reflections of the vehicle's surroundings are therefore simulated in the simulation.
  • the variations of the initial database occur by imposing changed probability distributions of traffic agents.
  • the probability distributions (such as normal distribution, binomial distribution, because Weibull distribution) are varied accordingly in order to design scenarios differently. For example, the value of a Standard deviation of random variables such as pedestrian behavior can be increased to obtain more differentiated scenarios.
  • Fig. 1 A method for designing and validating an automatic driving control system of a vehicle according to an exemplary embodiment of the invention.
  • SO data is provided from road users observed in reality with the help of a drone (pedestrians, cyclists, automated and non-automated cars, etc.) with assignment to the respective prevailing scenario in which the road users were observed, in particular one intersection, a highway, a highway drive, etc.;
  • the data of the road users observed by the drone also includes the trajectories of the road users, ie their trajectories with time information.
  • the current weather situation in the scenario under consideration is registered by the drone in order to reproduce it in the later simulations.
  • step S1 the data of the real road users recorded by the drone is processed to ensure the completeness of the data, so that the data that can be recorded by it is actually made available to the automatic driving control system of the vehicle, including its sensor units in the simulation environment, without being made available by the To capture data gaps arising from the drone.
  • the real road users captured by the drone are modeled as traffic agents for the simulation environment.
  • step S3 several simulations are carried out with the hybrid simulation environment generated with the initial database in multiple parallelization of the simulations and with a time scale factor greater than one for real time, integrating the automatic driving control system with simulation data instead of sensor data.
  • the simulation speed is significantly faster than real time, so that an enormous number of road users, scenarios and kilometers driven by the automated vehicle can be simulated with the automatic driving control system for designing and/or validating the driving control system in order to work with a quantity of data that has been obtained through practical tests (ie a real journey of the vehicle) would be practically unachievable.
  • variations of the initial database are carried out in order to use a sensitivity analysis to determine the respective reaction of the automatic driving control system to the variations in the hybrid simulation environment. Based on these variations and the results, a training data set and a validation data set are generated in order to single out particularly critical and safety-relevant situations.
  • an adaptation process S5 of parameters of a driving control function of the automatic driving control system can be carried out on the basis of the training data set. This particularly affects parameters of an artificial neural network that can be adjusted multiple times using supervised learning methods.
  • S6 a validation of the automatic driving control system is carried out based on the validation data set, with at least one algorithm of the driving control function of the automatic driving control system being validated.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Geometry (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Data Mining & Analysis (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Chemical & Material Sciences (AREA)
  • Mathematical Physics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Analytical Chemistry (AREA)
  • Computer Hardware Design (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Automation & Control Theory (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Robotics (AREA)
  • Traffic Control Systems (AREA)

Abstract

La présente invention concerne un procédé de conception et de validation d'un système de commande de conduite automatique. Les étapes consistent à : fournir (S0) les données des usagers de la route observés dans la réalité, avec leurs trajectoires, préparer (S1) des données et la modélisation des usagers de la route réels en tant qu'agents de circulation, créer (S2) un environnement de simulation hybride avec ajout de données synthétiques d'usagers de la route virtuels en tant que base de données initiale, réaliser (S3) des simulations dans une parallélisation multiple des simulations avec l'implication du système de conduite automatique à l'aide de données de simulation au lieu de données de capteur, des variations de la base de données initiale étant effectuées dans la simulation, générer (S4) un enregistrement de données d'apprentissage et un enregistrement de données de validation à partir de la base de données initiale modifiée, et réaliser (S5) un processus d'ajustement de paramètres d'une fonction de commande de conduite, et, à l'issue de tous les processus d'ajustement de la fonction de commande de conduite : exécuter (S6) une validation du système de commande de conduite automatique sur la base de l'enregistrement de données de validation, au moins un algorithme de la fonction de commande de conduite du système de commande de conduite automatique étant validé.
PCT/EP2023/060623 2022-06-22 2023-04-24 Conception d'un système de commande de conduite automatique dans des simulations massivement parallèles WO2023247092A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
DE102022206238.8A DE102022206238A1 (de) 2022-06-22 2022-06-22 Auslegen eines automatischen Fahrsteuersystems in massiv parallelen Simulationen
DE102022206238.8 2022-06-22

Publications (1)

Publication Number Publication Date
WO2023247092A1 true WO2023247092A1 (fr) 2023-12-28

Family

ID=86329133

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2023/060623 WO2023247092A1 (fr) 2022-06-22 2023-04-24 Conception d'un système de commande de conduite automatique dans des simulations massivement parallèles

Country Status (2)

Country Link
DE (1) DE102022206238A1 (fr)
WO (1) WO2023247092A1 (fr)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20210294944A1 (en) * 2020-03-19 2021-09-23 Nvidia Corporation Virtual environment scenarios and observers for autonomous machine applications
WO2021204983A1 (fr) 2020-04-09 2021-10-14 Siemens Aktiengesellschaft Dispositif de commande pour commander un système technique et procédé pour configurer le dispositif de commande

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE102017007136A1 (de) 2017-07-27 2019-01-31 Opel Automobile Gmbh Verfahren und Vorrichtung zum Trainieren selbstlernender Algorithmen für ein automatisiert fahrbares Fahrzeug
DE102019206908B4 (de) 2019-05-13 2022-02-17 Psa Automobiles Sa Verfahren zum Trainieren wenigstens eines Algorithmus für ein Steuergerät eines Kraftfahrzeugs, Computerprogrammprodukt, Kraftfahrzeug sowie System
DE102019219241A1 (de) 2019-12-10 2021-06-10 Psa Automobiles Sa Verfahren zum Erstellen eines Verkehrsteilnehmeralgorithmus zur Computersimulation von Verkehrsteilnehmern, Verfahren zum Trainieren wenigstens eines Algorithmus für ein Steuergerät eines Kraftfahrzeugs, Computerprogrammprodukt sowie Kraftfahrzeug

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20210294944A1 (en) * 2020-03-19 2021-09-23 Nvidia Corporation Virtual environment scenarios and observers for autonomous machine applications
WO2021204983A1 (fr) 2020-04-09 2021-10-14 Siemens Aktiengesellschaft Dispositif de commande pour commander un système technique et procédé pour configurer le dispositif de commande

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
BARBIER MATHIEU ET AL: "Classification of drivers manoeuvre for road intersection crossing with synthethic and real data", 2017 IEEE INTELLIGENT VEHICLES SYMPOSIUM (IV), IEEE, 11 June 2017 (2017-06-11), pages 224 - 230, XP033133712, DOI: 10.1109/IVS.2017.7995724 *
JASON BROWNLEE: "What is the Difference Between Test and Validation Datasets? - MachineLearningMastery.com", MACHINELEARNINGMASTERY.COM, 14 July 2017 (2017-07-14), pages 1 - 84, XP093063905, Retrieved from the Internet <URL:https%3A%2F%2Fmachinelearningmastery.com%2Fdifference-test-validation-datasets%2F> [retrieved on 20230713] *
NICO WEBER ET AL: "A Needle in a Haystack -- How to Derive Relevant Scenarios for Testing Automated Driving Systems in Urban Areas", ARXIV.ORG, CORNELL UNIVERSITY LIBRARY, 201 OLIN LIBRARY CORNELL UNIVERSITY ITHACA, NY 14853, 8 September 2021 (2021-09-08), XP091052412 *
W. WACHENFELDH. WINNER: "Autonomous Driving: Technical, Legal and Social Aspects", 2016, SPRINGER, article "The release of autonomous vehicles", pages: 425 - 449

Also Published As

Publication number Publication date
DE102022206238A1 (de) 2023-12-28

Similar Documents

Publication Publication Date Title
EP3970077B1 (fr) Procédé pour l&#39;entraînement d&#39;au moins un algorithme pour un appareil de commande d&#39;un véhicule automobile, produit de programme informatique, véhicule automobile ainsi que système
DE102018128290A1 (de) Verfahren und vorrichtung zum erzeugen von szenarien und parametrischen sweeps für die entwicklung und bewertung von autonomen antriebssystemen
DE102018128289B4 (de) Verfahren und vorrichtung für eine autonome systemleistung und zur einstufung
DE102017213634A1 (de) Verfahren und Vorrichtung für die Durchführung von virtuellen Tests in einer virtuellen Realitätsumgebung für ein autonom fahrendes Fahrzeug
DE102007053501A1 (de) Verfahren zur Entwicklung und/oder zum Testen wenigstens eines Sicherheits- und/oder Fahrerassistenzsystems für ein Kraftfahrzeug und Simulationsumgebung
DE102022112059B3 (de) Verfahren, System und Computerprogrammprodukt zur Kalibrierung und Validierung eines Fahrerassistenzsystems (ADAS) und/oder eines automatisierten Fahrsystems (ADS)
DE102020127107A1 (de) Verfahren und vorrichtung zum bestimmen einer konfiguration für ein autonomes fahrzeug
WO2020187563A1 (fr) Procédé pour l&#39;entraînement d&#39;au moins un algorithme pour un appareil de commande d&#39;un véhicule automobile, produit de programme informatique, véhicule automobile ainsi que système
DE102013200116A1 (de) Verfahren zum Entwickeln und/oder Testen eines Fahrerassistenzsystems
DE202022106107U1 (de) System zur Prüfung von automatisierten Fahrsystemen der Stufe 3 (ADS)
AT524822A1 (de) Verfahren zum Testen eines Fahrerassistenzsystems eines Fahrzeugs
DE102018207566A1 (de) System zum Durchführen von simulierten Kollisionsszenarios von einem Kraftfahrzeug mit einem nicht-motorisierten Verkehrsteilnehmer
DE102022127857A1 (de) Verfahren, System und Computerprogrammprodukt zur Sensitivitätsanalyse von Applikationsparametern für automatisierte Fahrfunktionen
DE102019212829A1 (de) Automatisierte Erkennung eines anormalen Verhaltens eines Verkehrsteilnehmers
WO2023247092A1 (fr) Conception d&#39;un système de commande de conduite automatique dans des simulations massivement parallèles
DE102019203205A1 (de) Verfahren zum Auswerten von Fahrzeugdaten sowie Fahrzeugdatenauswertesystem zum Durchführen eines derartigen Verfahrens
WO2022251890A1 (fr) Procédé et système pour tester un système d&#39;aide à la conduite d&#39;un véhicule
DE102021202083A1 (de) Computerimplementiertes Verfahren zum Trainieren wenigstens eines Algorithmus für eine Steuereinheit eines Kraftfahrzeugs, Computerprogrammprodukt, Steuereinheit sowie Kraftfahrzeug
DE102022203422A1 (de) Test einer automatischen Fahrsteuerfunktion mittels semi-realer Verkehrsdaten
DE102022107338B3 (de) Verfahren zum Testen von automatisierten Fahrzeugfunktionen
DE102022117841B4 (de) Verfahren, System und Computerprogrammprodukt zur Kalibrierung und Validierung eines Fahrerassistenzsystems (ADAS) und/oder eines automatisierten Fahrsystems (ADS) unter Berücksichtigung einer subjektiven Bewertung
DE102021214095A1 (de) Verfahren und System zum Erkennen von kritischen Verkehrsszenarien und/oder Verkehrssituationen
DE102022132917A1 (de) Verfahren und System zur Bestimmung der Kritikalität und Kontrollierbarkeit von Szenarien für automatisierte Fahrfunktionen
DE102023203633A1 (de) Hybrid-Verkehrssimulation mit Agentenmodellen zum Test teleoperierter Fahrzeuge
DE102023201782A1 (de) Limitierter Agentenmodelleinsatz in Hybrid-Verkehrssimulation mit Realtrajektorien

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23721888

Country of ref document: EP

Kind code of ref document: A1