ANRG

04/13/2021

Check out ANRG member Arvin Hekmati explaining his recent work on simulating COVID-19 spread in a university campus. This paper has been accepted to IEEE ICC'21 COVI-COM and can be found on arxiv at arxiv.org/abs/2104.04129 . This work shows that wearing masks reduces new cases by 3.6X and hybrid classes with 20% occupancy reduce new cases by an additional 2.15-2.3X. https://www.youtube.com/watch?v=oVQF8Vlw8FM

COVID-19 is now known to be spread through airborne transmission between individuals in an enclosed space even if they are far from each other. The goal of t...

03/12/2021

ANRG member Lilly Clark's paper "A Queue-Stabilizing Framework for Networked Multi-Robot Exploration" co-authored with NASA researcher Dr. Joseph Galante, and USC ECE Profs. Konstantinos Psounis and Bhaskar Krishnamachari has been accepted to the IEEE International Conference on Robotics and Automation and also appears in IEEE Robotic and Automation Letters. It presents a novel queue backpressure-based approach that would enable a network of robots to flexibly switch between different configurations over time to balance different needs such as connectivity, exploration, and relative localization. Perhaps, someday, teams of robots exploring Lunar caves or the surface of Mars may use such an algorithm! Check out the video below (https://youtu.be/xKM0vvIVHX8), and the paper online at https://ieeexplore.ieee.org/document/9361138

Motivated by planetary exploration, we consider the problem of deploying a network of mobile robots to explore an unknown environment and share information w...

09/21/2020

ANRG researchers won 1st place in a competition to develop a risk scoring methodology for COVID-19. Our approach tries to quantify the probability that an individual will get infected on a per-community basis, and represents that as a color-coded map. We demonstrated our ideas on LA County Public Health Data.

Recent data science challenge, hosted by the city, identifies a new, color-coded metric for individual risk.

05/10/2019

Kwame Wright with Bhaskar Krishnamachari after the PhD hooding ceremony yesterday. He has been working on distributed computing for connected vehicles.

06/12/2018

ANRG alumna Dr. Parisa Mansourifard has received a best paper award for the operations research track at the EU IEOM 2018 Conference to be held in Paris, France in July 2018. The paper, based on her Ph.D. dissertation research at USC is titled "A Game Theoretic Approach to Newsvendor Problems with Censored Markovian Demand." It offers an elegant closed-form solution to a novel robust min-max zero-sum repeated game formulation of the classic online decision making problem called the newsvendor problem, in which an adversary is allowed to choose customer arrivals at each time in a randomized manner from an uncertainty interval around the previous number of arriving customers, while the vendor tries to acquire the right number of newspapers to avoid over-stocking or under-stocking. While the formulation in this paper is mathematically abstract and rather general, it was originally inspired by problems in networks - for example, the problem of choosing a transport layer rate used by the TCP protocol on the Internet can be cast as such a problem.

06/07/2018

What are these blinking lights? A 50-node cluster of raspberry pi computers being used in our experimental work on distributed edge computing.

05/23/2018

Recent ANRG work on a new application layer client-server protocol that integrates streaming IoT data with micropayments using a cryptocurrency and transaction logging on a distributed ledger is described and featured in this excellent blog article..

What is it? And what does it mean for IOTA?

05/11/2018

ANRG members Shangxing, Pradipta and Pedro with Bhaskar after their hooding ceremony today.

03/09/2018

"Begin with the end in mind" goes a popular aphorism. In this recently posted peer-reviewed IEEE journal paper, ANRG researchers Nachikethas A. Jagadeesan and Bhaskar Krishnamachari apply this very principle to Radio-based localization (technology that can be used, for example, to track assets in an IoT-enabled smart warehouse or help people locate themselves via a mobile app in indoor malls). The Bayesian framework presented in this paper allows designers of location-tracking applications to specify a desired end-goal or objective for the localization system and it automatically optimizes the returned location estimates with respect to the specified objective. The framework generalizes and unifies prior work in the field.

We consider the problem of estimating an RF-device’s location based on observations, such as received signal strength, from a set of transmitters wi

03/06/2018

What else is new at ANRG? Our researchers, including PhD students Jason and Pradipta, as well as three undergrads, Yutong, Richard and Daniel have been hard at work developing a portable self-localizing low-power wireless robotics testbed called IRIS. The IRISbots use custom ultrasound hardware in combination with radio signaling to do ranging and location estimation using time difference of arrival. They can be programmed to do various interesting behaviors as a small swarm without any use of cameras.

This video illustrates an example application implemented on the IRIS testbed. In this example, three robots form a leader-follower chain and use the Time Di...

02/28/2018

What's new at ANRG? We recently released open source code for Jupiter, a novel system for dispersed computing (a generalization of grid computing, intended to handle more distributed and dynamic data-driven applications and dynamic networks). Jupiter automates the distributed scheduling and running of software that can be described by a computational task graph across a large number of (potentially geographically-distributed) computing nodes. It includes an implementation of WAVE, a novel framework for distributed task scheduling. As a sample dispersed computing application for testing, benchmarking and evaluation, we also developed and released a distributed network anomaly detection application (https://github.com/ANRGUSC/coded-dnad) that includes novel coded computing techniques developed and implemented by our colleagues Prof. Avestimehr, Annavaram and their groups, also from the USC EE department. Jupiter also includes a re-factored version of CIRCE, a task dispatch and run-time ex*****on environment for dispersed computing that we had originally developed and released last year as a stand-alone providing centralized scheduling. We have tested Jupiter on a 100-node cloud with nodes distributed across a dozen locations around the world; in principle it can handle even larger systems.

Jupiter is an Orchestrator for Dispersed Computing

09/21/2017

Cool new work from ANRG members Pradipta, Jason and Daniel (an undergrad). They developed ROMANO, an overlay protocol for robotic group communication that works on top of MQTT. Now you can command your robot army as a group and watch them respond in synchrony!

This demo shows how ROMANO protocol can be used for seamless control of a group of robots