The LE/FT lab performs research on applied information theory, emphasizing methods of probabilistic inference in Bayesian networks, factor graphs and related techniques. Our projects cover applications in digital communication, electronic circuits, and the interface between circuits and biology. The majority of our activities are centered on error correction codes and decoders, particularly algorithms for decoding low-density parity-check (LDPC) codes. Our projects also examine the broader relationship between energy and reliability in computing, communication and control systems. These topics increasingly overlap with biological systems, which provide models for efficiency and reliability that may be useful in electronic design.
Most of our activities revolve around two theoretical approaches deriving from information theory and cognitive science: (1) Bayesian approaches for circuit design using micro-power analog modules, and (2) Bayesian approaches using digital logic to filter highly random signals. These approaches are closely related, but can be manifested in a variety of application domains.
Here is a brief summary of our most recent research activities. Each activity can be regarded as an application domain for our Bayesian circuit perspective:
- Biocompatible wireless electronics: We are working to develop high-speed circuits for wireless data communication with bio-medical implants. We are especially focused on high-speed data links for brain-machine interfaces, which facilitate direct communication between electronic devices and cortical neurons. In this application, the key challenge is to deliver the maximum possible data rate without using too much power. Inside the body, power is converted into heat, which can damage surrounding tissues. Our current challenge is to deliver more than 100Mbits per second to an implantable receiver that consumes less than 10mW of power. Our solution involves the development of micro-power error-correction circuits and algorithms.
- Designing reliable behavior in genetic circuits: Engineers are now designing artificial genetic constructs that inserted into living organisms. These constructs, called "genetic circuits," can potentially be used for a myriad of applications, but the field is lacking in efficient procedures to specify, predict and verify the behavior of complex genetic circuits. We are studying the analogy that relates genetic circuits to noisy electronic circuits. By leveraging this analogy, we are discovering new techniques that may have benefits for both domains.
- Fault tolerant electronic systems: When electronic circuits are operated at the limits of their performance, they are increasingly susceptible to internal failures that may arise from electrical noise, from exogenous particle strikes, from interference with other devices, or from manufacturing defects. We are studying design approaches that can mask a very high rate of internal faults, yielding reliable operation as though no faults were present.
- Reliability and security in cyber-physical systems: Modern control systems rely on communication links between distributed resources that may include sensors, actuators and cooperating neighbors. We are interested in the ways these systems can be affected by communication reliability and quality-of-service. We are also studying new security vulnerabilities that are introduced by cooperative networked control systems. Our recent activity is specifically focused on applications for automated vehicle highways, where platooning strategies could theoretically be used to improve the safety, energy efficiency and capacity of our highways. To safely deploy these systems, we must first understand how to make them safe and secure in spite of unreliable communication links and the possibility of malicious attacks.