Publications

Real-time traffic monitoring has had widespread success via  crowd-sourced GPS data. While drivers benefit from this low-level,  low-latency road information, any high-level traffic data such as road  closures and accidents currently have very high latency as such systems  rely solely on human reporting. Increasing the detail and decreasing the  latency

Many emerging mobile applications, including augmented reality (AR) and  wearable cognitive assistance (WCA), aim to provide seamless user  interaction. However, the complexity of benchmarking these  human-in-the-loop applications limits reproducibility and makes  performance evaluation difficult. In this paper, we present EdgeDroid, a  benchmarking suite designed to reproducibly evaluate these  applications. Our

This paper shows how today's complex computing landscape can be  understood in simple terms through a 4-tier model. Each tier represents a  distinct and stable set of design constraints that dominate attention  at that tier. There are typically many alternative implementations of  hardware and software at each tier, but all

Edge computing processes data on infrastructure that is located close to the point of data creation. Mahadev Satyanarayanan recounts how recognition of the potential limitations of centralized, cloud-based processing led to this new approach to computing. Satyanarayanan, M. Nature Electronics, 2(1), January 2019

Real-time video analytics on small autonomous drones poses several  difficult challenges at the intersection of wireless bandwidth,  processing capacity, energy consumption, result accuracy, and timeliness  of results. In response to these challenges, we describe four  strategies to build an adaptive computer vision pipeline for search  tasks in domains such as

The generation of high-quality training data has become the key  bottleneck in the use of deep learning across many domains. We describe  Eureka, an interactive system that leverages edge computing and early  discard to greatly improve the productivity of experts in the  construction of a labeled data set. Our experimental

Recently, edge computing has emerged as a promising computing paradigm  to meet stringent quality-of-service requirements of an increasing  number of latency-sensitive applications. The core principle of edge  computing is to bring the capability of cloud computing in close  proximity to mobile devices, sensors, actuators, connected things and  end users, thereby

Wearable cognitive assistance applications can provide guidance for many facets of a user’s daily life. This thesis targets the enabling of a new genre of such applications that require both heavy computation and very low response time on inputs from mobile devices.  The core contribution of this thesis is

Accurate, up-to-date maps of transient traffic and hazards are  invaluable to drivers, city managers, and the emerging class of  self-driving vehicles. We present LiveMap, a scalable, automated system  for acquiring, curating, and disseminating detailed, continually-updated  road conditions in a region. LiveMap leverages in-vehicle cameras,  sensors, and processors to crowd-source hazard

VM handoff enables rapid and transparent placement changes to  executing code in edge computing use cases where the safety and  management attributes of VM encapsulation are important. This versatile  primitive offers the functionality of classic live migration but is  highly optimized for the edge. Over WAN bandwidths ranging from 5

An emerging class of interactive wearable cognitive assistance  applications is poised to become one of the key demonstrators of edge  computing infrastructure. In this paper, we design seven such  applications and evaluate their performance in terms of latency across a  range of edge computing configurations, mobile hardware, and wireless  networks,