Simulation as a Service for Cooperative Vehicles MASE 2019, Munich - - PowerPoint PPT Presentation

▶
simulation as a service for cooperative vehicles
SMART_READER_LITE
LIVE PREVIEW

Simulation as a Service for Cooperative Vehicles MASE 2019, Munich - - PowerPoint PPT Presentation

Nov 16, 2022 297 likes •448 views

Simulation as a Service for Cooperative Vehicles MASE 2019, Munich Jrg Christian Kirchhof, Evgeny Kusmenko, Bernhard Rumpe and Hengwen Zhang Software Engineering RWTH Aachen University http://www.se-rwth.de/ Motivation Our mission:

slide-1
SLIDE 1

Simulation as a Service for Cooperative Vehicles

MASE 2019, Munich Jörg Christian Kirchhof, Evgeny Kusmenko, Bernhard Rumpe and Hengwen Zhang Software Engineering RWTH Aachen University http://www.se-rwth.de/

slide-2
SLIDE 2

Jörg Christian Kirchhof | Software Engineering

2

Motivation

  • Our mission: provide MBSE methodologies for the automotive domain
  • EmbeddedMontiArc: Modeling language for automotive systems

Sensor 1 Sensor 2 Sensor N … Actuator 1 Actuator 2 Actuator M ECU 1 ECU 2 ECU K ECU 3

slide-3
SLIDE 3

Jörg Christian Kirchhof | Software Engineering

3

EmbeddedMontiArc

Deep Driving Component DPNet KalmanFilterBank

predictions: Q(-∞:∞)^{14}

Denormalizer

image:Z(0:255)^{3,210,280} affordance: Affordance

AffordanceToVector Localization DriverController

affordanceSmoothed: Affordance affordanceSmoothed: Affordance numLanes: N

SteeringBuffer

speedIn: Q brakingCmd: Q(-1:1) steerCmd: Q(1:1) timeRightIn: N timeLeft: N laneChange: LaneChange structure for affordance indicators predictions: Q(-∞:∞)^{14} timeLeftOut: N timeRightOut: N steerRecords: Q^{5} enum for lane change maneuvers accelCmd: Q(-1:1) laneChange: LaneChange

slide-4
SLIDE 4

Jörg Christian Kirchhof | Software Engineering

4

  • Cooperative Driving
  • Participant‘s decisions influence each other
  • Simulation of communication infrastructure

Requirements

  • Continuous Integration
  • Partition simulation to cloud computers
  • Frequently exchange parts of the code
  • Automatically evaluate simulation results
slide-5
SLIDE 5

Jörg Christian Kirchhof | Software Engineering

5

Architecture

(Sub-)Simulator Registry Controller Type-A register service discovery

Controller instances

requests Resource DB MQ Files API Server Map Service Controller Type-B Controller Type-C

RPC

(Sub-)Simulator Synchronization

slide-6
SLIDE 6

Jörg Christian Kirchhof | Software Engineering

6

Simulation Model

simulation Example1 { map = MunichCity.osm; startTime = 22.06.2017 13:30; deltaT = 1ms wheather = noRain, noSnow; duration = 30s; cars { AC-SE001:50°46'43.7"N 6°03'38.6"E ->

50°46'49.7"N 6°04'32.5"E,

M-SE003: … -> … }}

Sim

1 2 3 4 5 6 7 8 9 10

Start Position Target Position

"frames": [ { "deltaTime": 33, "position": [ 5.8806556, 50.83668824517681, 0.5429999999999999 ], "steering": 0.0165, "collision": false, ... "totalTime": 4059, "vehicleID": "45124“ }, ... ]

JSON

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

CI Pipeline

slide-7
SLIDE 7

Jörg Christian Kirchhof | Software Engineering

7

Sub-Simulator Synchronization

:Client

select/upload map, scenario

:API-Server

start the simulation reserve resource return resource

:Sub-Simulators :Registry

task setup(map, vehicle) execute simulation step release resource return result map sectoring execute simulation step execute simulation step return result return result return result synchronize synchronize

…

SD Sequence Diagram

slide-8
SLIDE 8

Jörg Christian Kirchhof | Software Engineering

8

Map Sectorization

slide-9
SLIDE 9

Jörg Christian Kirchhof | Software Engineering

9

  • 1. Sectorize map
  • 2. Find edges that go across sectors

(cut edges)

  • 3. In each sector find routes from one

cut node to the others (cut nodes)

  • 4. Find route between sectors

Route Planning

s t

Previous slide

slide-10
SLIDE 10

Jörg Christian Kirchhof | Software Engineering

10

Evaluation: CPU Consumption

Sector Overload

900 9000 90000 900000 1 5 10 15 20 25 30 35 40 45 50 Accumulated CPU time (ms) Number of sectors Total CPU Time Average CPU Time

slide-11
SLIDE 11

Jörg Christian Kirchhof | Software Engineering

11

Evaluation: Memory Consumption 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 1 5 10 15 20 25 30 35 40 45 50 Memory usage (Mbyte) Number of sectors Total Memory Average Memory

  • Max. Memory
slide-12
SLIDE 12

Jörg Christian Kirchhof | Software Engineering

12

Evaluation: Scalability 1000 2000 3000 4000 5000 6000 7000 8000 9000 5000 10000 15000 20000 25000 30000 35000 40000 45000 1 63 125 188 250 313 375 438 500 Total Memory usage (MByte) Total CPU time (ms) Number of vehicles Total CPU Time Total Memory

slide-13
SLIDE 13

Jörg Christian Kirchhof | Software Engineering

13

Evaluation: Simulation Speed 10000 20000 30000 40000 50000 60000 1023 4092 7161 10230 13299 16368 19437 22506 25575 Time in reality (ms) Simulator time (ms) 1 sector 2 sectors 3 sectors 4 sectors Online Simulation

slide-14
SLIDE 14

Jörg Christian Kirchhof | Software Engineering

14

Conclusion

(Sub-)Simulator Registry Controller Type-A register service discovery

Controller instances

requests

DB MQ

Files

Resource API Server Map Service Controller Type-B Controller Type-C

RPC

(Sub-)Simulator Synchronization

s t

"frames": [ { "deltaTime": 33, "position": [ 5.8806556, 50.83668824517681, 0.5429999999999999 ], "steering": 0.0165, "collision": false, ... "totalTime": 4059, "vehicleID": "45124"}, ... ]

JSON

1 2 3 4 5 6 7 8 9 10 11 12 13 14

simulation Example1 { map = MunichCity.osm; startTime = 22.06.2017 13:30; deltaT = 1ms wheather = noRain, noSnow; duration = 30s; cars { AC-SE001:50°46'43.7"N 6°03'38.6"E ->

50°46'49.7"N 6°04'32.5"E,

M-SE003: … -> … }}

Sim

1 2 3 4 5 6 7 8 9 10

Start Position Target Position

CI Pipeline