The
PHAROS (Petabit Highly-Agile Robust Optical System)
project funded under the DARPA’s CORONET
program provided an architecture, protocols and
algorithms for traffic engineering, resource management and signaling solutions
for highly-agile, large-capacity core optical networks. PHAROS main goals were rapid configuration
(less than 2 seconds), guaranteed data flow protection to up to 3 simultaneous
failures, high stability (including graceful degradation under excessive load),
and high degree of security/fault-tolerance in a multi-technology,
multi-vendor, and multi-domain environment. Key to achieving these goals was
PHAROS creation of abstract representations for all levels of the network. The
representations extend down to an abstract network model of the essential
contention structure of a node, and extend upward to address successive
(virtual) levels of functionality across the entire network. Thanks to these
multilevel topology abstractions PHAROS was able to achieve global
multi-dimensional optimization over the fundamental dimensions of network
management: network extent, technology levels, route protection, and
timescales. Abstraction allows a given request to be optimized across the
network, simultaneously trading off costs of resources within individual network
levels as well as the costs of transit between levels (such as the
optical-electrical boundary). Resources of all levels can be considered,
including wavelengths, timeslots, grooming ports, and IP capacity. With this
uniform approach, common to all levels of resource representation and
allocation, PHAROS accurately exploits the capabilities of all network
elements, while remaining independent of the switching technology or vendor
particularities. Another key enabler was PHAROS unitary approach that combines
the best features of centralized and distributed approaches. Long term planning
and coordination of resource usage (i.e., “Decision”) were undertaken by a
central entity (the Cross-layer Resource Allocator, or CRA), while the actual
switching elements’ configuration and quick reaction to failures (“action”)
were undertaken by a distributed network of control elements, following the
central entity guidelines (“playbooks”). PHAROS “centralized decision”
guaranteed stability and predictable dynamics, while PHAROS “distributed
action” achieved quick response to failures.
I was a key architect, and the lead of the CRA team.
International Technology Alliance (ITA) [2007-2009]
A US ARL- and UK MoD- led consortium of government, academy, and industry to perform joint research in network centric systems. My research focused on the capacity limits/laws for an ad hoc network of (secondary) cognitive radios in the presence of primary radios under Dynamic Spectrum Access rules.
Collaborative
Technology Alliance Program
(ARL/CTA) [2007-2009]
BBN participated in the
Communications and Networking (C&N) consortium of the Army Research Lab
(ARL) CTA program which has 5 different consortia: Advanced Sensors, Power
& Energy, Advanced Decision Architectures, Communications & Networks,
and Robotics. My research focused in the analysis of the impact of Multi-user
Detection (MUD)–capabilities in the performance of CSMA MACs and the design of
a simple yet efficient MUD-aware MAC.
The goal of this DARPA project is to build a large-scale MANET with very inexpensive nodes (by military standards), with up to four transceivers, each of which is highly frequency agile, and has a spectrum detector and simple MIMO capability. PIRANA will support multi-radio, multichannel dynamic spectrum access, unicast and multicast traditional and disruption tolerant routing and content based access. My research focused on HSLS extensions to scale up to one million nodes.
Stochastic Optimal Control Algorithms
and Next Generation Technologies for Dynamic Resource Allocation in Mobile
Communications Networks [2005-2006]
As a subcontractor to Scientific Systems Company, Inc.
(SSCI), BBN helped develop an integrated routing and scheduling protocol for
data transport in an opportunistic mobile wireless network using stochastic
control. We considered a very general mobile network communication problem in
which nodes have data to be transported to other nodes using randomly varying
channels subject to interference and dynamic frequency availability
constraints. The problem was formulated as a stochastic model predictive
control (MPC) problem in which a cost function of queue length is minimized.
The underlying computational problem is equivalent to a mixed integer linear
program where the integer variables arise from interference constraints. A
linear programming relaxation was implemented for the problem. The control
policy was evaluated using BBN's NeXt Generation (XG)
communication network simulation. The XG-simulation was a highly realistic
OPNET simulation developed originally under the DARPA XG program by BBN and
modified to include multi-hopping and to interface with MATLAB for controller
computations. The results showed that the developed system outperforms existing
techniques under all scenarios under consideration while providing stability
guarantees. The results suggest that improvements on the order of 100% are
possible when the traffic pattern consists of a single high speed flow.
This project was supported by the Army Research Office
(ARO), and I was the PI for the BBN effort.
The goal of this project was to accelerate the use of
software radios for wireless network research. As any frustrated
(mobile ad hoc) wireless network researcher can tell you, the lack
of flexibility in radio firmware severely limits experimentation with MAC layer
protocols. You have to live with what is in the radio (typically 802.11 for
most researchers). ADROIT attempted to change that by significantly enhancing
the open source GNU Radio software
to send/receive packets, control parameters and many basic radio functions,
except in software. The ADROIT system consists of the GNU USRP hardware,
RX and TX chain software, a MAC framework for easy development of MAC protocols
which is currently instantiated to a simple baseline, subnet layer routing
based on Hazy Sighted Link State routing, and the standard IP stack above it.
Unfortunately, this project did not run to its eventual completion for a number
of reasons unrelated to the project itself. However, some groups have picked up
whatever BBN did and are extending it. ADROIT was funded by DARPA IPTO and
included BBN (prime), Kansas U., MIT, UCLA, and Eric Blossom as team members.
Medium
Access Control for XG Communications (XMAC & XAP) [2003-2004]
DARPA ATO research
project. Developing an
architecture and protocol set for XG (Dynamic Spectrum Access) communication.
Joint Tactical
Radio System, Ground Mobile radios (JTRS GMR, formerly Cluster 1) [2002–2003]
Conducted a trade study
of existing routing protocols to determine their suitability for JTRS’s Wide
Band Networking Waveform (WNW), particularly the need to scale to 1600+ nodes. Since no existing routing protocols was
able to satisfy all WNW constraint, a new one (MALSR) had to be designed.
Besides being the co-inventor of MALSR – and analytically proving its
scalability to 1600+ nodes -- I was also a co-author of the scalable multicast
mesh algorithm used for sparse mode multicast traffic.
Utilizing
Directional Antennas for Ad Hoc Networks (UDAAN) [2001-2002]
DARPA ATO research project. Developed and implemented novel technologies to support and exploit beamforming antennas in ad hoc networks, including the first MAC protocol for beamforming antennas.
Density- and Asymmetry-adaptive Wireless Network (DAWN) [1999-2000]
DARPA ATO research project. Invented, analyzed, and implemented HSLS, the first ad hoc routing protocol that scale with network size. HSLS is an easy-to-implement link state variant that does not require complex hierarchies.