CCF - A Sparse Decomposition Framework for Complex System Design and Analysis
National Science Foundation – Urbashi Mitra (USC, lead PI), Antonio Ortega (USC, co-PI)
Cyberphysical systems necessitate significantly new design paradigms. They are immense in size and involve a complex interaction of communication networks, “natural” (environmental) networks as well as control. We propose to develop a novel theoretical framework to address these challenges by analyzing, tracking, and controlling Markov processes over graphs associated with complex systems. We shall exploit the correlation induced in these large Markov chains via sparse approximation theory employing tailored graph wavelets for representation. Next, we aim to design new sparse approximation algorithms and control methods tailored to the complex system. Finally, we propose to demonstrate the utility of the framework for system control via key application areas of interest such as cognitive radio, wireless body area sensing networks and potentially models for bacterial populations.
Active Communication, Sensing & Control in Actuated Underwater Sensing Networks
Office of Naval Research – Urbashi Mitra (USC, lead PI), Gaurav Sukhatme (USC, co-PI), Milica Stojanovic (Northeastern University, co-PI)
This project examines the design and analysis of deployments of autonomous nodes in the context of multi-scale, complex networks. We will focus on multi-vehicle, multi-target environments wherein the vehicles are enabled with multi-modal sensors, actuation and communication capabilities, but burdened by finite resources such as energy. These power limitations are in tension with the goal of large spatial coverage and long-term deployments. We shall continue to examine jointly optimized approaches wherein sensing (learning/classification), control/actuation and communication are coupled. In particular, novel machine learning methods for target localization and risk-aware planning for energy efficient motion will be undertaken. We directly tackle the problem of adjudicating between resource constraints and sensing gaps, partially known information, the challenges of achieving robust communication, and resource allocation.
Information Centric Networking on Wheels - Architecture and Protocols
Bhaskar Krishnamachari (USC, lead PI), Fan Bai (General Motors, co-PI)
This is a unique academia-industry collaborative project between researchers at the University of Southern California and General Motors. The primary goal of this project is to develop a hybrid network architecture for such vehicular networks which combines both the existing cellular infrastructure as well as new vehicle-to-vehicle (V2V) communication capabilities. The hypothesis is that such a hybrid network architecture will improve cost, capacity and robustness, compared to either a purely centralized cellular-based approach or a purely distributed V2V approach. Under a hybrid architecture, the project aims to design information-centric protocols for information dissemination, aggregation, and storage, that can exploit the spatio-temporally localized nature of vehicular applications. Further, through mathematical analysis, computer simulations, as well as experimental implementation on a research fleet of vehicles, this project aims to evaluate the performance of these protocols.
CPS: Breakthrough: Energy and Delay: Network Optimization in Cyber Physical Human Sensing Systems
National Science Foundation – Urbashi Mitra (USC, lead PI), Ashutosh Nayyar (USC, co-PI)
Wireless body area sensing networks (WBANs) can potentially revolutionize health care in the near term and enhance other application domains. The coupling of bio-sensors with a wireless infrastructure enables the real-time monitoring of an individual’s health, environment and related behaviors as well as the provision of real-time feedback with adaptive and personalized interventions. These WBANs represent a novel cyber-physical system that unites engineering systems, the natural world and human individuals. To develop the WBANs of the future, this project aims to investigate energy and delay sensitive sensing, communication, decision-making and control for health monitoring application of wireless body area networks.
Reduced-Dimension Wireless Network and Radio Design
Air Force Office of Scientific Research – Andrea Goldsmith (Stanford, lead PI) , Urbashi Mitra (USC, co-PI)
We propose to develop and analyze control methods for such complex, heterogeneous networks. In particular, we propose a framework which has the potential to seamlessly cope with the inherent system/information heterogeneity as well as dynamic network behavior by understanding the structure of the network and the associated optimized control in order to develop protocols and policies for high network performance. Specifically, we propose to employ sparse approximation theory to define meaningful methods for control and state estimation of complex networks by exploiting dynamical graph theory and multiscale methods which result in geometric representations of the network. Our proposed research has three, tightly interconnected thrusts: (1) Design and understanding of cognitive network control; (2) Development of compressed network control; and (3) Design and analysis of structured compressed sensing methods for network control and network state estimation.
Fundamentals of Wireless Communication Systems Using Orbital Angular Momenta
NSF – Andy Molisch (USC) and Alan Willner (USC)
One way to drastically improve spectral efficiency of wireless communications is to transmit multiple data streams together, using the same spectrum. However, in a “normal” setting, the receiver obtains then a mixture of those data streams, and they would interfere with each other. It is thus necessary to transmit the data streams in a way that allows the receiver to “disentangle” them. This project investigates a recently discovered such transmission method that is especially suitable for wireless links where the transmitter and receiver can “see” each other, and are fixed – such situations can occur, e.g., for wireless backhaul or wireless communication between servers in a data center. The method transmits different data streams on waves that have different orbital angular momenta (OAM), which describe the “phase twist” of a propagating wave. The current project investigates the fundamental science, as well as potential practical problems of OAM systems, and aims to assess their potential as a revolutionary method for drastically improving speed and efficiency of wireless data connections.
Crowdsourced Games for Multirobot Coordination
Nora Ayanian (USC, lead PI)
This project sets out an integrated research and educational approach toward establishing end-to-end solutions for distributed multi-robot coordination inferred from human coordination data crowdsourced from online multiplayer games. While the coordination methods of biological systems have inspired distributed multi-robot control in the past, the goal has often been loosely organized tasks such as flocking or nest-finding. Human coordination has been relatively unstudied in multi-robot systems, and presents a unique untapped opportunity for scientific discovery. This effort aims to develop new methodologies for multi-robot coordination firmly grounded in human collaboration, based on models learned from data collection via online multiplayer games. Specifically, the objectives are to 1) explicate the relationship between context (communication and sensing) and outcomes in distributed coordination games; 2) infer, using learning techniques, diverse coordination models for distributed teams of robots solving tightly coupled tasks; and 3) validate these models by evaluating their success in solving these tasks using a combination of simulation, hardware, and mixed reality experiments.
NeTS: Medium: Collaborative Research: Information Architectures for Femto-Aided Cellular Networks
Salman Avestimehr (USC, lead PI), Sanjay Shakkottai (UT Austin, co-PI), Ashu Sabharwal (Rice, co-PI)
Femto-aided cellular networks appear to be one the best solutions to achieve multi-fold capacity needed for future wireless networks. However, Femto-aided cellular systems have an information architecture that is very different from current planned and centrally managed cellular architecture. In this project, both the design of information architecture to acquire network state information and the optimal use of the resulting NSI will be addressed. The project is organized into three symbiotic research thrusts on network-aware physical-layer (PHY) coding schemes, network protocols and algorithms leveraging advanced PHYs, and their architectural prototypes. The first thrust utilizes recent innovations in deterministic models of wireless networks and develops novel physical-layer cooperative encoding and decoding schemes that operate with delayed, inconsistent, and erroneous NSI. The second thrust builds on the new physical-layer coding schemes to design network-scheduling algorithms to address performance issues. Finally, the third thrust utilizes the WARP programmable radios and studies implementation challenges of the protocols. The project goals of foundational design for Femto-aided cellular networks will have significant impact on industry practice. The PIs will facilitate technology transfer through their established industry affiliate program. A broad range of education and outreach activities will also complement the research agenda, including integration of research findings into the courses, promoting underrepresented and undergraduate populations, and engaging with the middle/high school community to raise the level of interest in engineering and mathematics.
EARS: Interference-Aware RF Theory and Design
Salman Avestimehr (USC, lead PI), Al Molnar (Cornell, co-PI)
The objective of this multidisciplinary program is to develop disruptive Radio Frequency (RF) technologies that provide significant spectral efficiency gains at the physical layer, by leveraging recent advances in physical layer interference management and integrated receiver design.
Dynamic Resource Allocation in Autonomous Multiagent Systems
Nora Ayanian (USC, lead PI)
This project will develop synergistic solutions to the modeling and algorithmic challenges of resource distribution among a team of robots for long-duration autonomy. The goals of this project are to overcome the algorithmic and modeling challenges of long-duration deployment for teams of autonomous vehicles. Specifically, the work will focus on the development of a synergistic modeling and deployment framework, using reuse closed loop supply chain and deployment strategies which deliver an automatically synthesized concurrent solution for coordination, planning, and control for the delivery of resources via autonomous vehicles. The objectives of this proposal are to develop a new modeling and coordination framework for UXVs: (1) to recover and distribute reusable resources such as energy and sensors to enable robots to operate beyond their on-board energy or sensor capabilities in degradative environments; (2) to enable human-in-the-loop decision making for risk assessment, priority adjustments, and in-field requests; and (3) that is experimentally validated via a proof-of-concept heterogeneous team of unmanned ground and aerial vehicles. The novelty of the work lies in the interdependent approach to resource allocation modeling and deployment for long-duration autonomy.
Measuring and Mitigating the Impact of Network Bias on Computation in Graphs
Kristina Lerman (USC, lead PI)
Graphs are often massive, making it infeasible for graph mining algorithms to obtain a global view of the graph and data on it. To analyze such graphs, scalable graph-mining computation relies on local algorithms, such as belief propagation, that aggregate data locally in a neighborhood of some vertex and propagate results to neighboring vertices. Successive iterations integrate these values to improve results of distributed computation. However, graph structure can dramatically impact performance of distributed algorithms by introducing systematic differences between local and global views of the graph and the data on it. These differences may bias results of local computation on graphs. Proposed work will quantify this local bias effect. The work will explain how this bias depends on the structure of the graph, develop a theoretical framework to quantify it for any graph, as well as computational approaches to mitigate the bias.
Stochastic Dynamic Optimization & Games: Simulation and Learning Methods
Rahul Jain (USC, lead PI)
In this project, we have developed theory and algorithms for optimal decision-making and control of distributed, non-linear, stochastic dynamical systems. Such systems are ubiquitous in the military, and are expected to be even more so in the future. Examples are robots that operate autonomously in combat zones providing defensive and offensive support in missions, platoons of under-water vehicles that cooperatively patrol the sea-ways for potential mines, and to detect enemy activity, and squadrons of unmanned, autonomous UAVs that patrol the skies for air domination. Design of such systems throws up new problems and resurfaces old challenges in stochastic programming, mathematical control and game theory. One must contend with three types of challenges in such systems: (i) optimization under probabilistic/risk constraints, (ii) sequential optimization over a long horizon (possibly without any model), and (iii) multi-agent optimization and coordination.
Load Balancing by Network Curvature Control
Edmond Jonckheere (USC, lead PI), Francis Bonahon (USC, co-PI)), Bhaskar Krishnamachari (USC, co-PI)
This project explores the interplay between the topology/geometry of networks and their traffic load pattern with the ultimate objective of deriving new load-balancing algorithms based on curvature control. Among other activities, the project has lead to the design of a novel variant of queue-aware routing called Heat Diffusion Routing and its implementation and evaluation on the USC Tutornet low power wireless IoT testbed. Ongoing activities are exploring curvature concepts relevant to scheduling in multi-channel multi-hop wireless networks.
Smarter Markets for a Smarter Grid: Pricing Randomness, Flexibility and Risk
Rahul Jain (USC, lead PI), Abhishek Gupta (Ohio State, co-PI)
Electricity infrastructure is under-going a gradual but accelerating evolution. This is throwing up new challenges to manage uncertainty on a scale never seen before in any setting. Current methods of managing uncertainty via adequate spinning reserves are uneconomical. New market designs and pricing mechanisms have to be investigated that facilitate renewable energy integration and demand side management in the smart grid while ensuring sufficient grid stability margins. In fact, the very nature of renewable energy as a “random” good can be regarded as a novel concept. Moreover, requiring flexibility on both the demand and generation side itself requires pricing incentives to elicit the maximum flexibility possible. Risk management is a well understood topic. But building a risk-awareness itself in the market design is a new notion. Designing markets for the next generation grid, which will integrate a lot more renewable energy and will also be a lot “smarter” requires a smarter design for the markets.