Wireless Networks

Society is rapidly approaching a pervasive computing environment in which a variety of devices communicate constantly with each other, organizing themselves, gathering information from the environment, and responding by actuating a range of robotic devices. Key to the flexibility and effectiveness of these systems is the ability to communicate without the limits imposed by phone lines, cables, or cell phone towers, but ad hoc wireless communications impose new challenges on researchers, including bandwidth constraints, increased errors, the need for self-organization, and the energy limits inherent in many deployed sensor and actuator systems. Members of Rensselaer’s Center for Pervasive Computing are working on all levels of the technology needed, from new physical devices to improved routing of messages.
 

High-Speed Optical Networks
 

With funding from the National Science Foundation (NSF), Shivkumar Kalyanaraman, associate professor, and Partha Dutta, assistant professor, both of electrical, computer, and systems engineering (ECSE), are using inexpensive LEDs (light-emitting diodes) and microelectronics techniques to create a multihop wireless system with the capacity to bring broadband optical connections to homes. They have developed techniques through which the nodes can discover that they’re aligned and then pass messages back and forth. Dr. Kalyanaraman is also working with students in Rensselaer’s undergraduate research program on a nearer-term project to bring broadband wireless capacity to homes in an area within a mile of the Rensselaer campus. This project initially uses Pringles cans to create radio frequency (RF) antennas, but it eventually will include both RF and optical components as well as GPS positioning and new routing software. The goal is to demonstrate an inexpensive system, which can be upgraded as technology advances, with the goal of uniting communities around the world.

Once a wireless community network is formed, a group including Dr. Kalyanaraman has Army funding to investigate ways to prioritize messages sent using the popular IEEE 802.11 MAC (media access control) protocol. They are developing joint source network coding that will route high-priority messages over the fastest routes, while using slower routes for less important messages. In a complementary project, ECSE Professor and Department Chair Kenneth Connor and Adjunct Professor Ted Anderson are developing smart plasma antennas, similar to fluorescent tubes or neon signs. Since they disappear when not in use, they could be particularly valuable for military applications.

Wireless Networks

A number of Rensselaer researchers are developing methods to make wireless networks more reliable and efficient, including:

Failure Predictions ECSE Professor Kenneth Vastola and former graduate student Lisa Shay, now a professor at West Point, created a wireless communications system that can recognize when a link is about to fail. The system is designed for situations in which small mobile units or individual soldiers must communicate on ad hoc wireless systems, with their movements sometimes taking them out of range or behind a building or hill. The system predicts such failures as much as 30 to 40 seconds in advance, giving time to alert superiors and get instructions. 
 

Compression to Minimize Costs When sending multimedia messages over ad hoc wireless networks, packets of information should be compressed to reduce congestion and transmission times along the route. But compressing the messages makes them more sensitive to channel impairments, as losses of a single packet can render the entire message unreadable. Since packets of information travel different distances, with some passing through highly congested routes and others processed by less-occupied nodes, one must vary the degree and mode of compression and error correction coding, depending upon the network path and the channel conditions along that path. Power consumption is also an important trade-off. Alhussein Abouzeid, ECSE assistant professor, and William Pearlman, ECSE professor and director of the Center for Image Processing Research, are working on a project known as COCO, which tries to optimize computation and communication costs in multimedia wireless networks. They are developing power-aware algorithms, based on RPI’s award-winning image compression algorithm, SPIHT (Set Partitioning in Hierarchical Trees), to detect and manage congestion and power consumption by jointly optimizing compression and error correction coding.
 

Packet Routing Costas Busch, assistant professor of computer science, studies packet routing problems and is interested in a concept of fairness in which all users are satisfied. He is working on power-aware reversal routing in mobile ad hoc networks, in which nodes that are overburdened by messages or that lose power reverse themselves, forcing packets to take another route. With Bulent Yener, associate professor of computer science, he is also working on methods to synchronize nodes in wireless networks to minimize packet collisions and save energy.
 

Choosing the Best Channels Gary Saulnier, ECSE associate professor, is developing coding to help messages being sent on wireless networks make the most efficient use of physical resources such as antennas. He is creating a system to optimize transmissions when a group of users depends on a particular suite of carriers. The self-learning system probes channels to determine if they are already in use and chooses those most able to carry the messages.
 

Distributed Sensor Networks and Robotics

A large group of Rensselaer researchers works on distributed networks of sensors and actuators,

devices to learn about the environment and to react in specific ways. Key challenges include limited power, the requirement for real-time information, and the need for systems that can self-organize and reconfigure themselves. A few of their projects include:

Distributed computer Vision With support from an NSF CAREER award, Richard Radke, ECSE assistant professor, is developing distributed algorithms for networks of cameras to gather and exchange information for potential applications including terrain mapping, object tracking, query-by-image-example, and view synthesis. Cameras dropped over a battlefield, for example, have to figure out where they are in relation to each other, a task Dr. Radke approaches by having them identify and exchange information about unique features. Such a camera network must self-calibrate without central command-and-control and be robust to dynamic, unreliable communication channels and low-power contraints.

 
Robots that work well together Wesley Huang, assistant professor of computer science, is affiliated with Rensselaer’s state- and industry-supported Center for Automation Technologies (CAT). His research field is motion planning in distributed robotic systems. He is interested in ways groups of autonomous mobile robots can coordinate among themselves to accomplish tasks for such jobs as hazardous waste cleanup, search and rescue, security patrolling, or reconnaissance. Jeff Trinkle, professor and chair of computer science, develops mathematical methods to model dynamic systems with discontinuities such as robots that have unpredictable contact with their environment. He is interested in systems such as dense nets of mobile sensors that could enter a collapsed building, map it, and help determine the safest way to disassemble the structure to rescue those trapped within.

 
Environmental Monitoring Rensselaer administers the Upper Hudson Satellite Center of the new Rivers & Estuaries Center on the Hudson, a research and education center supported by state and federal governments, not-for-profit organizations, and other funding. A key portion of Rensselaer’s mission is to develop monitoring, communication, and visualization tools to analyze the Hudson, using both moored sensor systems and autonomous underwater vehicles. Sandra Nierzwicki-Bauer, director of the DFWI, professor of biology, and a key collaborator on the Hudson River Project.

Contacts: