Course Info

Schedule: Friday, 12:45pm - 15.45pm, room: L-2710 (Lassonde building)

Instructor & contact: Jerome Le Ny, Associate Professor, Electrical Engineering department

Office Hours: after class, or by appointment, or stop by my office to see if I'm available

Discussion forum: Piazza

Announcements

• 01/12/18: First class

Course Description

The "Internet of Things" refers to large networks of devices, such as sensors, actuators and computing devices, from microcontrollers and smartphones to mainframes and data centers, which interface the physical and cyber world, can communicate with with each other and often interact closely with humans. Application examples include smart buildings, smart energy grids, and even smart cities, consumer wearables and intelligent medical devices, or connected cars. Engineers have already been connecting physical devices through local or even large networks, but typically for specialized industrial applications and in controlled environments. The current trends are toward a much higher level of interconnections, deploying massive sensor and actuator networks outside of factories and around humans to enable smart infrastructures, integration with the standard internet of computers and mobiles, leveraging cloud computing platforms, etc. This raises many fundamental questions, from design (how to organize these networks, how and where to handle computations) to purpose (how do we make these connected machines more useful) to security (physical infrastructures now exposed to cyberattacks) and privacy (think "Big Brother is watching you"). The purpose of this course is to explore these issues and what the future of automation might look like. We introduce some modelling techniques to formally analyse these “Cyber-Physical Systems” (CPS) and present tools that can be used to design and deploy useful, reliable, safe and secure CPS, as well as limit their intrusiveness.

Topics:

• Networked and embedded control systems (NECS)
• Decentralized control of multi-agent systems (MA)
• Technological evolutions from industrial SCADA systems to IoT and cloud computing (DCS)
• Security and privacy in large scale monitoring and control systems (DCS)

Course Philosophy, Prerequisites, and Grading

The course emphasis is on breadth rather than depth. That is, for most topics we will typically not discuss the absolute state of the art, and you will not necessarily be ready to start developing new results immediately. For some topics (ex: decentralized control), we will just give a very basic introduction via selected problems of interest. Indeed, many of the lectures could be worth developping further into a full course. However, this course will give you an idea of some types of issues that are currently being actively investigated by control theorists and engineers (ex: typical session topics at automation conferences). In addition, I will introduce during our discussions a number of analysis, design and computational tools that are broadly relevant and useful in systems and control, but not necessarily part of the standard curriculum yet. The target audience I have in mind is: MS or beginning PhD students in controls/automation looking for a research topic or interested in broadening their perspectives and knowledge of the field, and possibly senior undergraduate honors students specializing in this area. Given the small size of the class, I'll be happy to work closely with you while you learn, so feel free to stop by my office or after class to ask questions about the material, show me your current homework and project results, chat about other ideas, etc.

Prerequisites: previous exposure to dynamical system modeling and control, mathematical maturity, and a motivation to learn and explore further on your own as needed. This is a course focusing on models and fundamental theory for CPS applications, not on the engineering aspects of building (the necessary) computing and communication infrastructures. However, these aspects should by definition influence the models and abstractions of CPS that we study, so we will need to discuss them to some extent (in simplified ways), understand and manipulate toolkits, etc.

Grading: homework (40%), 15 minute mini-lecture and short accompanying notes on a technical paper (20%), final project (40%). The problem sets explore the class topics in more depth and generally require a computer. Students will also be responsible for teaching a mini-lecture, on a topic determined with the instructor, which includes the preparation of a few pages of lecture notes (this will be adjusted based on the number of participants). The project can be done alone or in pairs (recommended), exceptionally in groups of 3 (with permission of the instructor). It can be for example:

1. Exploring in some details a topic by studying a paper or group of related papers, performing analysis and simulations on a reasonably complex original case study, leveraging available software tools, etc. For example, you could try to design a digital control system to illustrate one or several artifacts due to modern implementation platforms, with the goal of being as realistic as possible.
2. Further exploring implementation issues and IoT software toolkits for distributed control systems. This could include hardware set-ups (physical implementation is not required; note that the course does not discuss embedded systems programming). Evaluate the effect of network delays, what is appropriate for edge vs. cloud computing, etc.
3. A serious attempt at a novel contribution on a topic related to this class, which has the potential for example to lead to a publication (for advanced students, perhaps already doing research in this area).

Auditors: Auditors are welcome in the class. It is likely that student auditors will be required to give a class presentation as well.

References

NECS/Embedded Computing: Alur (2015), E. A. Lee and Seshia (2017), Hristu-Varsakelis and Levine (2005), Zhang, Branicky, and Phillips (2001), Hespanha, Naghshtabrizi, and Xu (2007), Mirkin (2007), Hetel et al. (2017), Heemels, Johansson, and Tabuada (2012)

Multi-Agent Systems: Bullo, Cortés, and Martínez (2009), Mesbahi and Egerstedt (2010), Bullo (2017)

ICS/SCADA/DCS/IoT: Galloway and Hancke (2013), Geng (2017)

Cloud/Parallel/Distributed Computing: GoogleU (2011), Armbrust et al. (2010)

Fault Detection & Security in CPS: Blanke et al. (2016), Slay and Miller (2007), Sundaram (2012), Pasqualetti, Dörfler, and Bullo (2013)

Privacy: Dwork and Roth (2014), Le Ny and Pappas (2014)

Fundamentals in control theory: Åström and Murray (2016) (basic concepts), Vidyasagar (2002), Jönsson (2001) (input/output modeling), Isidori (2017) (more advanced), van der Schaft (2017) (dissipativity), Åström and Wittenmark (1997) (sampled-data systems), Chen and Francis (1995) (sampled-data systems, more advanced), Scherer and Weiland (2015) (LMIs, dissipativity), Boyd et al. (1994) (more LMIs)

Some other courses on related topics: ASU, UPenn, Georgia Tech, Lund, UWaterloo, EdX: Data Science Ethics