You are on page 1of 21

Next Gen Embedded Control SW Development and Validation: Challenges & Solutions

S. Ramesh GM R&D (ramesh.s@gm.com)

GM Confidential

Electronics & Control Software


On the increase in several domains Its complexity is exploding Performing several critical functions

GM Confidential

Electronics, Control & Software shifting the basis of competition in vehicles


- More functions & features in Software
100M Lines of Code (+9900%)

Fuel Cell Wheel Motor

Value from Electronics & Software

Hybrid PT Software $ 13%

Software $ Other $ 9% 2% Electronics $ 13% OnStar OBD II HI Spd Data

ACC Rear Vision


$1182 (+196%) 50 ECUs (+150%)

Electric Brake Other $ DoD 8% GDI

BCM
Electric Ignition ... ABS

Mechanical $ Rear aud/vid TCC


$400

Passive Entry Software $ Electronics Side Airbags $ Mechanical $ Head Airbags Other
1M Lines 20 ECUs

Electronics $ 24% Mechanical $

EGR Electric Fan 76% 1980s

CDs 1990s

55%
AVG.

1970s

AVG.

2000s

2010s

2020s

ABS: Antilock Brake System ACC: Adaptive Cruise Control BCM: Body Control Module DoD: Displacement On Demand ECS: Electronics, Controls, and Software Subsystem Controls & Features

Source: Matt Tsien, GM

EGR: Exhaust Gas Recirculation. Vehicle Integration GDI: Gas Direct Injection OBD: Onboard Diagnostics System Connection TCC: Torque Converter Clutch PT: Powertrain Forefront of Innovation GM Confidential

GM Confidential

GM Confidential

GM Confidential

GM Confidential

GM Confidential

Active Safety to Autonomy

Wheel Motor Brake by Wire

(Partial Steer-by-wire)

EPS+AFS

By-wire applications Partial and Fully Autonomous Federated to Integrated functions

Hybrid Power Train

V2I CAN 360o Sensing

V2V

GM Confidential

Central Problems

How do we develop such large, complex and critical software that is correct and provides the necessary confidence? Are the current processes, methods and tools for V&V adequate?

What are the possible solutions and what challenges for their implementations?

GM Confidential

10

SW Development

Human-intensive
Considered a creative experience Predominantly use of mental models and artifacts Non standard Quality human dependent

Requirements

Huge Gap between Requirement and Code


Requirements vague and informal Code is formal

Human errors natural, inherent and have low

traceability

1 bug in every 40 50 lines of code (Industry estimates) Errors lead to poor quality, high costs, conservative designs extreme human hardship and even serious safety issues

Code

GM Confidential

11

GM Confidential

12

Model-based Development

Requirements

Executable models used as intermediate step


Simulink/SF, UML

f() fcg 1 onOff 2 accelResume 3 cancel 4 decelSet 5 brake 6 gas


speed gas dSpeed brake decelSet cancel activ e accelResume onOf f throttleDelta trigger

1 active

Code auto-generated Gap Still exists between Requirements and Design Models V&V focused only on code

2 throttleDelta

CruiseMain
CruiseMDL

3 speed
throttleDelta activ e speed inactiv eThrottleDelta drag

7 8 drag inactiveThrottleDelta

Plant

Code

GM Confidential

13

V cycle of Development and V&V


Simulation and Analysis Requirements (System, Functional) Test Benches, Test Vehicles Vehicle Validation

System Level

Functional Architecture Development Physical Architecture Development

Functional Integration

System Validation

Component Level

Implementation, & Unit Testing

Integrate SW & HW (Component)

Focus on Integration testing (Domain & Vehicle level testing) Many artifacts are informal and ambiguous Test generation manual
GM Confidential 14

Fallout

Significant efforts for Verification and Validation with the explosion of Software Functionality
Automation of Verification Steps Early Verification catching bugs as they enter

Higher standard of quality in Development and Verification of SW


Rigorous Processes, Methods and Tools

GM Confidential

15

Formal Methods integrated in Model Based Development


Formal Requirements models Transition systems Math functions Formal logic Rigorous analysis Consistency Correctness Completeness

Requirements
Design models
Transition Systems Hybrid automata

Design

Design Verification Timing Analysis Model checking Distributed Task schedules

Automatic code generation


Code generator verification Translation Validation

Implementation

Formal Verification Automatic Synthesis

Model-based testing
Test generation using model checking Better and effective coverage
GM Confidential

Testing
16

Formal Methods alone not enough


Formal methods is one (probably small) part in the solution Software Engineering practices need to change
Good Requirement Engineering Standardized and proven Components and execution infrastruture Product-line approach to development Rigorous Change management processes and Tools

GM Confidential

17

Technical Challenges

Scalability of formal methods


Striking the right level of abstraction Earlier in the life cycle the better the scalability, the more robust the results

(Time and Space) Distributed Development of Applications


Compositional verification methods

Extreme dependence on manual steps


Less/poor use of documentation Lack of traceability Non standard activities

Focus on product-line rather than individual products Lack of inadequate data for automatic V&V

GM Confidential

18

Techno Political Challenges

Existing internal `proven-in-use practices


Resistance to experimentation and exploration

External Tool vendors influence


Mathworks, IBM, Reactis

Changing SW Platforms and Standards


CAN vs Flexray vs Ethernet Autosar vs Home-grown platforms

Mounting Resource Constraints


Struggling to complete yesterdays work no question of tomorrows technologies

GM Confidential

19

Conclusions

SW for Next Gen Embedded Systems pose several challenges Automation and Rigorous Techniques are key requirements for meeting the challenges of Verification and Validation
Formal methods employ mathematical artifacts that provide the necessary rigor Formal artifacts are precisely analyzable with automatic support Formal methods enhance quality and can lead to aggressive designs

SW Engg. Practices need to change significantly


Move away from human dependent steps More standardization and less freedom for tomorrows programmers The chasm between requirements and design to be bridged

GM Confidential

20

Thank You

GM Confidential

You might also like