REAL AND VIRTUAL PROVING OF AUTOMATED DRIVING IN BERLIN'S MIXED TRAFFIC Dr. Ilja Radusch, ilja.radusch@fokus.fraunhofer.de 10.05.2017
WHO AM I? Director Smart Mobility at Fraunhofer Fraunhofer is Europe largest applied R&D society Technology development and consulting Connected and automated driving Security, trust, identity management for vehicle networks Analysis of large mobility data sets (big data) Parking and parking space surveillance Conducted Europe largest field operational test on connected driving 2
WHO AM II? Director of Daimler Center for Automotive IT Innovations Pre-competitive R&D applying the latest information technology to the automotive domain in both product and process innovations Fostering tri-lateral cooperation between automotive industry and Daimler AG Success stories include prototyping the Remote Parking Assistant of the current Mercedes E-Class 3
AUTOMATED DRIVING Industry pushing for following scenarios till 2020 Scenario 1: Structured and simple environment Highway traffic jam chauffeur till 60 km/h Highway chauffeur till 130 km/h Scenario 2: Complex but still structured environment in parking garages Parking chauffeur will allow driver less maneuvers and parking in environments mixed with non-automated vehicles and pedestrians 4
PROVING AUTOMATED DRIVING IS TIME CONSUMING Roads & Infrastructure Automotive Cloud 240 million kilometers or 1.000 vehicle life times just for a single vehicle Connected Vehicles Automated Vehicles 5
PROVING AUTOMATED DRIVING IS AN ONGOING PROCESS Roads & Infrastructure Automotive Cloud While development to validate your code! During deployment to validate your data! Connected Vehicles Automated Vehicles 6
PROVING AUTOMATED DRIVING IS AN ONGOING PROCESS Roads & Infrastructure Combining real and virtual proving Automotive Cloud Connected Vehicles Automated Vehicles 7
BENEFITS OF VIRTUAL PROVING Virtual Proving consists of either replay of pre-recorded data real-time simulation of vehicles and sensors in relevant traffic scenarios Time and cost effective solution Allows to simulate vehicles and sensors for Quick validation during development Input data generation Virtual validation of whole system with estimated improvements in algorithms or sensors 8
TEST FIELD FOR AUTOMATED DRIVING IN BERLIN Area will be equipped with sensors for ground truth collection Support for vehicle-2-x communication, including traffic lights Connected and automated driving functions Situation analysis and prediction HD maps for robust self-localization 9
INDEPENDENCE OF ENVIRONMENTAL CONDITIONS 10
OUR APPROACH EVEN WORKS INSIDE GARAGES 11
DYNAMIC COUPLING OF BEST-IN-CLASS SIMULATORS Vehicle Simulation LIDAR, vehicle dynamics, in-vehicle sensors, User Simulation driving behaviour, emotions, mood, Traffic Simulation vehicle flow, traffic lights, real traffic patterns, Virtual Proving with Simulation ISAAC Robot Simulator Communication Simulation ITS G5 ad-hoc, LTE, 5G ADAS Functions platooning, self-driving, 12
DECOUPLING ALGORITHM AND DATA GROOMING Virtual proving environment allows for quick synthetic data generation Passive data Vehicle probe data, from vehicle fleets, xfcd probe data, from smart phones, busses, Infrastructure traffic data, from infrastructure operators, Quality Active data generated dynamically through interconnected, best-ofclass simulators Synthetic data allows testing of algorithms independent of probe data quality, feature set, and quantity Signal set Quantity 13
REAL AND VIRTUAL PROVING TESTBED Comprehensive virtual proving testbed VSimRTI Integrates best-in-class simulators Already bundled with extendable LIDAR, vehicle, traffic, and communication simulators Research and commercial licenses offered Developed for over 10 years, used by over 140 institutions all over the world Makes development of self-driving functions more time and cost effective Puts several vehicles and sensors in traffic scenarios for (regression) testing during development Active data generation Apply for trial license simulation@dcaiti.com 14