Ashish Kashinath

Dept. of Computer Science

University of Illinois at Urbana-Champaign
Urbana, IL 61801-2307

Office: SyNeRCyS Lab, Coordinated Science Laboratory (CSL) 440
Phone: (858) 291-2161;
Email: ashishk3@illinois.edu

I am Ashish, a Systems PhD student working on Cyber-Physical Systems (CPS), advised by Prof. Sibin Mohan.

At a high level, my work looks at Predictable Networking and Fault tolerance for CPS made out of commodity components. Towards this, some of my work involves designing networks for CPS, that are predictable and deterministic (hence safer), resilient to failures and cheaper (using commodity protocols and hardware). Examples of domains where my research is useful include : Automotive Networks, Industrial Control Systems, Power Grid (Substation) Networks and Aerospace Networks. Prior to Illinois, I obtained Masters and Bachelors degrees from UC San Diego (California, US) and NIT/REC Tiruchirappalli (Tamil Nadu, India).

Checkout other projects from our group at projects page.

Thanks to Prof. Sibin's philosophy to working on identifying problems in systems research seen in practice, I have had the opporunity to collaborate with other industry practioners and faculty. I have interned at Aviatrix Systems (U of I Research Park), VMware (xLabs/Research Group), NVIDIA (GeForceNow Systems Group) and Microsoft Research (Hardware, Devices & Cyber-Physical Systems Group). Some of my collaborators include faculty at UC San Diego (Prof. Rajesh Gupta + MESL Lab), Intel Labs (Dr. Javier-Perez Ramirez) and the University of Utah (Dr. Luis Garcia).


Research

Current Projects:

Analytical Models for Hard Real-time Cyber-Physical Networks

In this work, I am developing further analytical (yet practical) models of networks for Cyber-Physical systems. The goal is to overcome limitations of current models that are conservative models based on Network Calculus and Empirical Measurements, which results in loss of guarantees when scaled up to networks having either more flows or switches. I am currently looking at models inspired by feasible region calculus and combining it with standard techniques in performant modern networks such as Software-defined Networking (SDN)


Resilient Networks using Communication-Control co-design

In this work, I am looking at approaches to capture physical and computational effects into the network model


Modeling Safety in PLC NetworksIn this work, I am looking at discrepanies between real and simulated models of PLC networks and characterizing the safety of the networks


Past Projects:

RealFlow : A System for Designing Commodity Cyber-Physical Networks

This work develops an end-to-end system using only commodity platforms and protocols for design and deployment of CPS networks. CPS networks- though self-contained and small in scale, have higher stringent guarantees due to safety as well as certification requirements. Current techniques for developing CPS networks are domain-specific, proprietary and are not performant. RealFlow is a tool comprising of : A) analytical, correct-by-construction framework that gives safety guarantees and B) deployment framework for using the results from the analytical framework on Pica-8 SDN switches. Thus, we can use RealFlow to design and deploy Cyber-Physical Networks using only commodity platforms and protocols, making it cheaper.


PIRMedic : Building Dependable Occupancy Sensors for Smart Buildings

This work studies how Occupancy Sensors based on Passive Infra-Red (PIR) sensing fail and studies techniques to detect and diagnose such failures. The work focused on characterizing the phenomenon of Pyrolectric Effect and capturing behavior of Voltages and Currents on the sensor hardware platform at the Edge to detect and diagnose failures. Note that this is in contrast to classical data-driven anomaly detection and fault diagnosis techniques which makes this work novel. This work takes a step forward in improving the quality of data in Smart Buildings, which can then be used upstream by building automation pipelines. Higher quality of data from sensors can pave way for meaningful inference and reduced operational costs when IoT is deployed on a large scale in Buildings.



Academic Papers: White Papers, Short Papers, Work-in-Progress and Posters:

Mentoring

I spend sometime every week mentoring students in research and computing in general. I work with the following wonderful students:

Parvati Menon and Waidat Bada

Parvati and Waidat did a wonderful job bringing up our CPS-based network environment with Raspberry Pi's, NUCs and Industrial LogicSupply Computers. They are evaluating and setting up various usecases from the industry and translating them into a lab environment within our lab in addition to working closely with the Engineering IT to setup the network with appropriate isolation and efficiency requirements


Natalie Gavin

Natalie focused on analyzing and studying PLC applications by using the OpenPLC environment. She was quick to learn the different way of interfacing a water valve sensor to both an Arduino Microcontroller and a Raspberry Pi as well as bringing up the OpenPLC runtime.


Simon Xia

Simon is a sophomore student working with me via the URAP program.


Chen-Wei ChouChen-Wei is a junior with interest in Real-time OS and Systems Programming.



Other Links

TeX & other writing-related commands I refer often: Useful Commands
Class I assisted, as a TA at UC San Diego or Illinois: Graduate Computer Networks Computer Organization & Systems Programming Real-time & Cyber-Physical Systems Database Systems
Social Media: I try to avoid social media. I still I am there on it though.
Service: JSys student Editorial Board
Miscellaneous: I like running through nature - through the corn fields here in Illinois and along the La Jolla Shores beach in San Diego. I have run the Triton 5k, a run through the scenic campus of UC San Diego multiple times including a couple times after graduating - 2016, 2017, 2019 and 2021. I am yet to run a half marathon - but actively working towards it.