Pat Pannuto

Assistant Professor, Computer Science Engineering
University of California, San Diego | 3202 CSE | +1 858 822 2924 | Calendar

Publications | CV | Talks | Teaching | Fellowships & Awards

Headshot photo of Pat

I am interested in the boundary between the digital and physical world. My research aims to expand the reach of computational infrastructure to sense and actuate more of the physical world.

My expertise is in the design and implementation of resource constrained computing systems. These are systems whose deployment in the world constrains their form factor, connectivity, deployability, and maintainability, which often must then operate on microwatts of power, with only kilobytes of memory and effectively kilobit/second lossy communication links – yet the systems as a whole must be accurate, timely, and reliable.

I am always excited to learn about new and interesting problems that we may be able to solve. In the past, this has included efforts to embed computation in more places with the world's smallest computer, to build systems for human social interaction tracking to support epidemiologists and psychologists, to design infrastructure to support city-scale sensing, and to develop technology capable of country-wide grid health monitoring.

Selected Honors & Awards


A history of my teaching prior to joining UCSD is also available.

Michigan Micro Mote & MBus

The Michigan Micro Mote, or M3 project, aims to bring general-purpose computing and sensing to millimeter-scale devices. As part of this effort, we identified the system interconnect as a key impediment to further scaling the energy and area of embedded computing. To address this, I led the design of MBus, a new chip-to-chip interconnect optimized for energy-conscious designs.

The Michigan Micro Mote is on display in World's Smallest Computer exhibit the lobby of the Computer History Musuem in Mountain View. You can see the exhibit online here.

Cubeworks is commercializing the M3 technology.

M3 node on a finger

M3 node on edge of US nickel coin

{ JSTS'16 | Micro Top Picks '16 | ISCA'15 | WARP'15 | CICC'14 | VLSI'14 | VLSI'14 | JSSC'13 }

The Tock Operating System

Low-power microcontrollers lack some of the hardware features and memory resources that traditionally enable multiprogrammable systems. Accordingly, microcontroller-based operating systems have not provided important features like fault isolation, dynamic memory allocation, and flexible concurrency. However, an emerging class of embedded applications are software platforms, rather than single purpose devices, and need these multiprogramming features. Tock, a new operating system for low-power platforms, takes advantage of limited hardware-protection mechanisms as well as the type-safety features of the Rust programming language to provide a multiprogramming environment for microcontrollers.
Tock Logo
Tock isolates software faults, provides memory protection, and efficiently manages memory for dynamic application workloads written in any language. It achieves this while retaining the dependability requirements of long-running applications.

Google recently announced the OpenSK, a fully open-source hardware security module, which is built on top of Tock.

{ SOSP'17 | APSys'17 | PLOS'15 }


Location information is a key aspect of context-aware computing. Making location a first-class computational resource, like time synchronization or networking today, is critical to realizing visions of intelligent and reactive environments.

I have worked on several localization efforts, tackling different applications:


For most of modern history, massive effort has gone into the careful organization, sorting, and filing of information for later retrieval. With the rise of the Information Age, we have transitioned from filing information to simply searching for it on demand, expecting our computing infrastructure to automatically find exactly what we are searching for. The same is not yet true for the physical world. While there are exceptions, the vast majority of things in the physical world remain invisible to the computational domain. Slocalization takes one small step towards enabling "search not file" for the physical world, demonstrating a sub-microwatt tag that can be localized with decimeter accuracy in complex, indoor environments using ultra wideband backscatter.
Channel impulse response after repeated integrations

Frontside of tag hardware

Backside of tag hardware

{ IPSN'18 }


In late 2013, DecaWave released the first (and as of 2019 still only) commercially available ultra wideband transceiver. While ultra wideband provides the potential for extremely accurate range estimates, in practice single range estimates can exhibit variation well over a meter. SurePoint explores what's required to build a scalable, high-fidelity, and high-reliability (29 cm 50th percentile, 77 cm 99th percentile accuracy of raw range estimates) system atop a UWB ranging primitive.
{ SenSys'16 | HotWireless'15 }


Ultra wideband tracking (order 1-10 cm) with only narrowband components of lightweight (3 g), low power (75 mW or 3.9 mJ/fix), low cost (< $5 USD), fast-moving (up to 2.4 m/s) tags.
Harmonium tag mounted on a micro quadrotor
{ TOSN'18 | IPSN'16 | MC2R'15 | HotWireless'14 }


An exploration in using lighting infrastructure for astral navigation indoors. Luxapose spawned considerable follow-on work (with which I have no affiliation) that is really cool and looks at how to do this without requiring intelligent LED lighting infrastructure, in particular check out LiTell and some of Xinyu's other light-based localization work.
Luxapose image processing snapshot
{ MobiCom'14 }


An exploration in relative location, a low-power (126 J/day) wearable badge that provides high fidelity (centimeter-accurate, second-level granularity) human interaction information.

Opo was initially developed in partnership with epidemiologists interested in understanding the impact of interaction time and distance on the spread of flu in elementary school settings. The platform has since also been adapted to support handwashing studies in hospital settings.

Various configurations of people wearing Opo tags
{ SenSys'14 }

GridWatch and The Open INcentive Kit

The power grid is one of humanity's most significant engineering undertakings, and it is essential in developed and developing nations alike. Yet, most grids have remarkably little introspection into their operation. GridWatch is inspired by a simple observation, a smartphone stops charging for one of two reasons: (1) The user unplugged it, (2) The power went out. Given that we can filter the first case with the phone's accelerometer, can a critical mass of smartphones act as low-fidelity, high-coverage, fine-granularity grid monitor that runs independent of local utilities? Following the promise shown by this initial inspiration, GridWatch has grown to include a rich suite of power sensing technologies, which includes custom hardware for high-fidelity measurements and ground truth and most recently ideas for non-contact voltage monitoring from project collaborators.

Today, GridWatch has micro deployments in the United States, Venezuela, Nigeria, and India. The primary deployment, branded DumsorWatch, in Accra, Ghana serves as one of the principle measurement and evaluation arms for the $498 million Ghana Power Compact and has several hundred sensors deployed and several thousand app-based participants.

Born out of the complex, wide-area study demanded by DumsorWatch, the Open INcentive Kit (OINK) is a new platform for running and managing incentive-based studies. With OINK, experimenters set up a series of rules that express when and how study participants should be incentivized. OINK monitors a study, automatically triggers incentives, and handles disbursements.

{ COMPASS'19 | GetMobile'19 | ICTD'19 | MobiCom'18 | HotMobile'14 }

The Signpost City-Scale Sensing Project

City-scale sensing holds the promise of enabling deeper understanding of our urban environments. However, a city-scale deployment requires physical installation, power management, and communications—all challenging tasks standing between a good idea and a realized one. The Signpost project aims to provide a platform that enables easy deployment and experimentation for city-scale applications.


A deployed signpost in the wild
{ IPSN'18 }


These are the conferences I generally follow, courtesy of ConfSearch.

Working on, or reviewing, any empirical work? Check out the Empirical Evaluation Guidelines checklist.

I am happy to share slides from any presentation I give. I generally try to post slides from major talks, but if something is missing that you are interested in, please reach out and I will add it.

I have been fortunate to be successful in many fellowship applications. This is due in no small part to excellent guidance from my colleagues and mentors. I am happy to share some of that advice and my application materials here.

Here are my application materials from the faculty market. Hopefully they are helpful.

One of my favorite parts of academia is the opportunity it has afforded me to travel the world. Here's all the countries I've made it to thus far (also US states I've spent any appreciable time in).

One lesson from travel is that much of the web is inaccessible from areas where internet access is unreliable or expensive. There is a neat tool that estimates the literal cost of loading your webpage. I appreciate Dan Luu's metric of aiming to be smaller than the JavaScript used by Google's AMP.

I am a proud alumnus of the Lab11 research group, headed by Dr. Prabal Dutta. During graduate school, I was a National Defense Science & Engineering Graduate (NDSEG) Fellow (2013), a National Science Foundation Graduate Research Fellowship Program (NSF GRFP) Fellow (2013), and a Qualcomm Innovation Fellowship (QInF) Fellow (2013). My graduate studies were also generously supported by the Semiconductor Research Corporation through both the TerraSwarm program (part of STARnet) and the CONIX center (part of JUMP).