Control as a Service (CaaS)

Transcription

1 Control as a Service (CaaS) Cloud-based Software Architecture for Automotive Control Applications Hasan Esen*, Hideaki Tanaka +, Akihito Iwai # DENSO (*Europe, + Japan, # Silicon Valley) Masakazu Adachi TOYOTA CENTRAL R&D LABS., INC. Jens Knodel, Dominik Rost, Christian Peper Fraunhofer IESE, Germany Alberto Bemporad, Daniele Bernardini ODYS Srl, Italy , SEC Workshop, Seattle, US

2 CONTENTS 2 I. Automotive Trends II. CaaS: Moving the Controllers to the Cloud? III. Case Study IV. Conclusion

3 CONTENTS 3 I. Automotive Trends II. CaaS: Moving the Controllers to the Cloud? III. Case Study IV. Conclusion

4 1. Automotive Trends Vehicle System Trend ~60 s 90 s 2010 s 2030 s 4 Stand Alone Control Networked Control Integrated Vehicle System (IVS) IVS + Infrastructure Increasing number of Electronic Control Units (ECUs) ECU program size SW complexity

5 1. Automotive Trends 5 Propulsion Technology in % Advanced Diesel (1) Diversified Mobility Smart Grid in 2030 Diesel hybrid Advanced flexible combustion techniques Advanced spark ignition Spark ignition hybrid 0% Plug-in hybrid range ext. Full Electric (inc. Fuel Cel) Adapted ICE Resource: (1)ERTRAC/EUCAR) Mega Competition

6 1. Automotive Trends 6 Automated Driving (AD) and Communication Technologies in 2030 Remote Control Center AD AD V2I V2V System of Systems Increased Complexity & Scale: Safety, Security, Reliability Problems

7 CONTENTS 7 I. Automotive Trends II. CaaS: Moving the Controllers to the Cloud? III. Case Study IV. Conclusion

8 CaaS: Moving the Controllers to the Cloud? 8 TODAY data domain is moving to the cyber space Telematics Cyber world Physical world control domain Cruise Transmission Engine data domain Pre-crash Brake Steering Sensor/Actuator Microcontroller

9 What can be more? 9 POTENTIAL FUTURE controllers will also move to the cyber space? Telematics Engine Motor Pre-crash Cyber world Steering Brake Cruise Physical world control domain data domain Sensor/Actuator Safety Microcontroller

10 Potential Future: Moving the Controllers to the Cloud 10 Idea: Control as a Service (CaaS) Engine Motor Pre-crash Steering Brake Cruise

11 Selected Scenarios 11

12 Concept Platform and Software Architecture 12 Architecture-Centric Engineering Solutions (ACES) Development of a concept prototype Partner: Dr. Jens Knodel

13 First Tests 13 Setting Throttle Control Variants (Virt. ECU) Round-Trip Times Cloud Server Location: Kaiserslautern, Germany

14 Challenges 14 Too Many! SOFTWARE Efficient computing Real-time cloud technology Reliable/fault-tolerant SWarchitecture COMMUNICATION hard real-time control-aware (wireless) communication Security, safety, privacy CONTROL Sensor fusion Communication-aware control Role distribution: Cloud vecus/in-vehicle ECUs

15 CONTENTS 15 I. Automotive Trends II. CaaS: Moving the Controllers to the Cloud? III. Case Study IV. Conclusion

16 4. Case Study 16 CONTROL Sensor fusion Communication-aware control Role distribution: Cloud vecus/invehicle ECUs

17 Case Study Scenario 17 Adaptive Cruise Control with Active Steering in a Cloud-Controlled Vehicle Assumptions: The vehicle has CAN-Bus and a WLAN gateway for cloud communication. Obstacles are detected by a distance sensor Objective: Achieve torque control from cloud coupled with active steering for obstacle avoidance taking WLAN latencies into account If possible, brake and adjust to leader velocity otherwise, steer. Proposed Solution Methodology: Stochastic Model Predictive Control Partner: Prof. Alberto Bemporad

18 Case Study Model Predictive Control (MPC) Design 18 I. Nominal Deterministic MPC Standard MPC; no delay compensation subject to (s.t.) constraints C 1 Find optimal torque rate (DT) minimizing the cost J s.t. C 1 II. Network-aware Stochastic MPC Markov transitions + delay compensation J tot S i 1 p i J i p i : probability of scenario i J i : cost function for scenario i S: # of scenarios (2) subject to constraints, where i = 1,2 Find optimal torque rate (DT) minimizing the cost J tot s.t. C i

19 Case Study Results 19 Simulation conditions Obstacle detection sensor range: 30m in front of the car Time-varying reference velocity: 30km/h to 50km/h Two obstacles: the first can be avoided by braking, the second requires steering 1 delay step in idle, 3 delay steps in busy Probability to switch Markov state is 10% (both idle to busy and busy to idle) Reference velocity 50 km/h 2nd Obstacle 30 km/h 1st Obstacle time

20 Case Study Results 20 Performance assessment I. Nominal Deterministic MPC Standard MPC; no delay compensation II. Network-aware Stochastic MPC Markov transitions + delay compensation

21 CONTENTS 21 I. Automotive Trends II. CaaS: Moving the Controllers to the Cloud? III. Case Study IV. Conclusion

22 Conclusion 22 Emerging mobility trends remote traffic observation and vehicle control diverse inter & intra vehicle communication networks automated driving platooning vehicle to grid integration powertrain electrification alternative fuels Automotive systems: more complex, more connected, larger in scale Safety, security, reliability and privacy problems: very challenging to solve Interdisciplinary, advanced solutions; new methodologies are required

23 Conclusion 23 On-going work related to CaaS Platform COMMUNICATION Security threat analysis SOFTWARE Cooperative CaaS Architecture CONTROL Optimal use of communication medium Event-Based Control (EBC)