File preview
CPS Open Source Tools and Experimental Platforms for CPS
Birds of a Feather Scribe: G. Karsai
BOF Guidelines
“The BoF sessions are intended to be seeds for future special interest groups in the CPS Virtual Organization. Please help this community “selforganization” by stimulating discussions on
a. interesting and substantial challenge problems that can guide research, b. shared testbeds and experimental platforms, c. recommendations toward funding agencies, d. key industry stakeholders”
Open Source Tools and Frameworks Introduction - Motivation
Serves as: Experimental and repeatable validation of research
Experimental testbed/s Challenge problems (not necessarily Grand…) Peer-reviewed open source tools E.g. http://www.escherinstitute.org/Plone/toolqualification To industry, small business, etc.
Serves as: Means of publishing research results
Serves as: Vehicle for tech transfer
Open Source Tools and Frameworks Discussion
Researchers already do it in some areas…but
How to handle improvements? CPS Sourceforge? (open collaborative environment for development) Need: stable open architecture
Maintaining quality and longevity
Stable data interface in tools Processor emulators, network emulators – stable platforms
QC needs: quality rating, degree of independent validation
Open Source Tools and Frameworks Discussion
Need: Data sets for experimentation
E.g.: Vehicle data, driver behavior data E.g.: Challenge problems documented
Challenge problems: need to be quality controlled, vetted Who pays for the experimental platform
Physical platform? – Some industrial solutions already Small platforms already exists – but larger vehicles are different
Need something common -- a platform
Open Source Tools and Frameworks Discussion
Support: From Office of Cyber Infrastructure?
Long-lifetime code Community used code
Open source license is required in DOEsponsored research Problems:
Maintaining versioned community resource requires effort Meta-data: Self-describing systems : meta-data comes with the component (e.g. to assist with tool integration)
Open Source Tools and Frameworks Discussion
Evaluating research – (Esp. young faculty)
Peer-reviewed/evaluated research code Start from a working system in CPS, replace/ improve domain-specific parts (controls, scheduling, software, etc.) and show improvement AirStar (NASA) Starmac (C. Tomlin, UCB) Robocup simulation
Platform examples:
Open Source Tools and Frameworks Discussion – To summarize
Goals:
Scientific results, reproducibility Productivity of (grad) researchers, accessibility of results Common platform, architecture
To stay honest: competition! NSF should facilitate the process
Birds of a Feather Scribe: G. Karsai
BOF Guidelines
“The BoF sessions are intended to be seeds for future special interest groups in the CPS Virtual Organization. Please help this community “selforganization” by stimulating discussions on
a. interesting and substantial challenge problems that can guide research, b. shared testbeds and experimental platforms, c. recommendations toward funding agencies, d. key industry stakeholders”
Open Source Tools and Frameworks Introduction - Motivation
Serves as: Experimental and repeatable validation of research
Experimental testbed/s Challenge problems (not necessarily Grand…) Peer-reviewed open source tools E.g. http://www.escherinstitute.org/Plone/toolqualification To industry, small business, etc.
Serves as: Means of publishing research results
Serves as: Vehicle for tech transfer
Open Source Tools and Frameworks Discussion
Researchers already do it in some areas…but
How to handle improvements? CPS Sourceforge? (open collaborative environment for development) Need: stable open architecture
Maintaining quality and longevity
Stable data interface in tools Processor emulators, network emulators – stable platforms
QC needs: quality rating, degree of independent validation
Open Source Tools and Frameworks Discussion
Need: Data sets for experimentation
E.g.: Vehicle data, driver behavior data E.g.: Challenge problems documented
Challenge problems: need to be quality controlled, vetted Who pays for the experimental platform
Physical platform? – Some industrial solutions already Small platforms already exists – but larger vehicles are different
Need something common -- a platform
Open Source Tools and Frameworks Discussion
Support: From Office of Cyber Infrastructure?
Long-lifetime code Community used code
Open source license is required in DOEsponsored research Problems:
Maintaining versioned community resource requires effort Meta-data: Self-describing systems : meta-data comes with the component (e.g. to assist with tool integration)
Open Source Tools and Frameworks Discussion
Evaluating research – (Esp. young faculty)
Peer-reviewed/evaluated research code Start from a working system in CPS, replace/ improve domain-specific parts (controls, scheduling, software, etc.) and show improvement AirStar (NASA) Starmac (C. Tomlin, UCB) Robocup simulation
Platform examples:
Open Source Tools and Frameworks Discussion – To summarize
Goals:
Scientific results, reproducibility Productivity of (grad) researchers, accessibility of results Common platform, architecture
To stay honest: competition! NSF should facilitate the process