DYNA4 with DYNAanimation in Co-Simulation with SUMO vehicle under test Virtual testing by coupling high fidelity vehicle simulation with microscopic traffic flow simulation Dr.-Ing. Jakob Kaths TESIS GmbH 1 st 3D Mapping Solutions User Conference TESIS GmbH www.tesis.de
Motivation Generation of complex traffic scenarios is time consuming. Variation of traffic scenarios is not straight-forward, because of deterministic behavior. scene from DYNAanimation with deterministic DYNA4 traffic Co-Simulation of the virtual vehicle (DYNA4) in virtual traffic (SUMO) to overcome these shortcomings Reproducing real traffic situations (e.g. A9 Holledau, Friday 5pm) or including realistic traffic control is difficult. 2
Agenda Basics of vehicle dynamics vs. microscopic traffic flow simulation Co-Simulation of DYNA4 with SUMO 1 Example applications Automated driving on highways Object detection in a complex urban scenario Summary 1 http://sumo.dlr.de/ 3
# of stops Basics of vehicle dynamics vs. microscopic traffic flow simulation (I) virtual vehicle Simulink-based DYNA4 virtual traffic open-source SUMO virtual vehicle with high fidelity driving dynamics, MBS axles, drivetrain and sensor models hundreds of vehicles with humanlike car following and lane changing behavior vehicular measurements: yaw rate, wheel rotation, microscopic traffic measurements: number of stops, travel times, direct integration of vehicular control systems from MiL to HiL for ECU tests SiL integration of traffic signal controllers 4
Basics of vehicle dynamics vs. microscopic traffic flow simulation (II) virtual vehicle Simulink-based DYNA4 testing in deterministic scenarios for accurate reproducibility virtual traffic open-source SUMO stochastic traffic scenarios to account for variation in real world direct support of OpenDRIVE road format included conversion from OpenDRIVE format Unity-based 3D-animation, e.g. for testing of object detection simple 2D-animation for verification purposes the virtual vehicle in the virtual traffic Enables testing of vehicular systems in complex, stochastic and yet reproducible traffic scenarios. 5
Native support of OpenDRIVE in DYNA4 DYNA4 offers a native support of OpenDRIVE files No conversion into proprietary road format On-the-fly road generation with automated terrain creation for generic and fast scenario generation Shown example: digital A9 motorway test bed created by 3DMapping Solutions including road geometry, lane markings, guard rails, traffic signs, 6
Co-Simulation - DYNA4 s part DYNA4 offers a high fidelity model for the vehicle under test (VuT) additional deterministic surrounding traffic VuT and deterministic traffic objects known in SUMO driving on OpenDRIVE road network deterministic high-performance subscription to traffic objects surrounding VuT with optional semi-circle extension full availability of DYNA4 features including Simulink-based sensors and GPU-based sensors by looping SUMO traffic through DYNA4 traffic modular integration of SUMO by using C++ TraCI API within Simulink S-Function SUMO VuT +DYNAanimation 7
Co-Simulation SUMO s part SUMO 1 delivers consistent traffic scenarios and is open-source hundreds of vehicles, bicycles and pedestrians back-propagation of spillbacks closed-loop reaction to traffic control including actuated traffic control immediate availability of basic information e.g. (vehicle type, position, speed, yaw angle) information enrichment necessary (e.g. vehicle model, wheel base) realistic behavior with regard to traffic phenomena is desired, but requires reasonable input data for traffic signal control and demands 1 http://sumo.dlr.de/ screenshots from SUMO/NETEDIT 8
Co-Simulation of DYNA4 and SUMO S-Function C++ no SUMOstep? set VuT yes position, speed, angle VuT SUMO step (10Hz) DYNA4 step (1000Hz) get traffic information traffic objects left? yes interpolate traffic information no position, speed, angle traffic signal timings VuT: Vehicle under Test 9
Setting up a scenario for Co-Simulation of DYNA4 and SUMO OpenDRIVE road DYNAanimation vehicle parameters vehicle control DYNA4 Co-Simulation SUMO traffic demand priority rules & signal timings online offline engineering tasks 10
Co-Simulation of DYNA4 and SUMO Vehicle under Test 11
Automated driving on motorway A9 for testing of AEB deterministic DYNA4 traffic Driving on OpenDRIVE road (Digital Motorway Test Bed A9) Seamless combination of surrounding SUMO traffic and deterministic DYNA4 traffic Shown example: testing of autonomous emergency brake with deterministic vehicle cutting in 12
Automated driving on motorway A9 in different traffic states Driving on OpenDRIVE road (Digital Motorway Test Bed A9) in different traffic states Simple variation of traffic demand, available lanes, etc. Shown example: lane closure due to broken down truck 13
Automated driving on motorway A9 in different environmental conditions Driving on OpenDRIVE road (Digital Motorway Test Bed A9) Simple variation of environmental conditions such as precipitation (rain, snow), weather (sunny, overcast) etc. 14
Object detection in complex urban scenario Camera-based object detection tested in DYNAanimation Probabilistic behavior of detection algorithms requires repeated runs with varying scenarios Shown example: Complex traffic scenario with object detection trained with COCO 1 dataset 1 http://cocodataset.org/ 15
Summary Co-Simulation of the virtual vehicle from DYNA4 in virtual traffic from SUMO Overcoming step size gap of both tools Using OpenDRIVE as mutual road network Seamless integration of deterministic and stochastic traffic Easy variation of traffic states and environmental conditions Complex traffic scenarios including vehicles, pedestrians, animals, bicycles and traffic signals One-click scenario variation for stochastic, but reproducible testing Facilitate frontloading for ADAS/AD functions by virtual testing in complex traffic scenarios. 16
Enhance virtual test drives in DYNA4 with traffic flow simulation for testing and development in complex traffic scenarios. Thank you for your attention! For more information visit us at our booth! Dr.-Ing. Jakob Kaths e-mail: jakob.kaths@tesis.de phone: +49 89 74 7377-63 TESIS GmbH www.tesis.de