Research Overview

From MurrayWiki
Revision as of 04:53, 23 July 2012 by Murray (Talk | contribs)

Jump to: navigation, search

This page contains a brief summary of my group's current research activities, broken up into the three main areas. More information is available on the individual project pages below and also in the recent publications from my group.

Analysis and Design of Biomolecular Feedback Systems

Bfs-overview.png

Feedback systems are a central part of natural biological systems and an important tool for engineering biocircuits that behave in a predictable fashion. The figure at the right gives a brief overview of the approach we are taking to both synthetic and systems biology. There are three main elements to our research:

  • Modeling and analysis - we are working to develop rigorous tools for analyzing the phenotype of complex biomolecular systems based on data-driven models. We are particularly interested in systems involving feedback, since causal reasoning often fails in these systems due to the interaction of multiple components and pathways. Work in this are includes system identification, theory for understanding the role of feedback, and methods for building and analyzing models built using high-throughput datasets.
  • In vitro testbeds - we are making use of both transcriptional expression systems and protein expression systems to develop "biomolecular breadboards" that can be used to characterize the behavior of circuits in a systematic fashion as part of the design process.
  • Biocircuit design - engineered biological circuits required a combination of system-level principles, circuit-level design and device technologies in order to allow systematic design of robust systems. We are working on developing new device technologies for fast feedback as well as methods for combining multiple feedback mechanisms to provide robust operation in a variety of contexts. Our goal is to participate in the development of systematic methods for biocircuit design that allow us to overcome current limitations in device complexity for synthetic biocircuits.

Current projects:

Recent papers:

Architectures for Control Using Slow Computing

Slowncs-arch.png

Current techniques for the design of software-enabled control systems rely on the existence of high performance sensing, actuation and computational devices that can be embedded within a physical system at modest cost. The goal of this project lies at the other end of the computational spectrum: we seek to develop new principles and tools for the design of closed loop control systems using highly distributed, but slow, computational elements. Our motivation is to develop new architectures for feedback control systems that can be used in applications where computational power is extremely limited, such as systems for which the energy usage of the system must remain small. Current thrusts include:

  • Control design using time-delay: Control design using slow computing requires that we design controllers that have time delays that are significant compared to the underlying dynamics of the process we seek to control. By using combinations of (pre-computed) feedforward computation and combinations of differentially delayed feedback signals, we seek to design feedback systems that exploit time delays rather than avoid them.
  • Design of information flows: Networked control systems include complicated interconnections between subsystems, with nonlinearities and time delays as integral elements of the system model. How do we analyze stability and performance of this class of systems in a way that exploits the structure of the interactions? How do we design the information flows and other elements of the system to obtain desired behavior in the presence of uncertainty?
  • Bio-plausible control systems: One set of applications for control systems using slow computing are on bio-inspired and bio-engineered systems. In both cases, we seek to develop control approaches that are compatible with the computations that are possible in neurons and/or biomolecular systems, including understanding the role of underlying stochasticity in our computations.

Current projects:

Recent papers:

Design and Verification of Protocol-Based Networked Control Systems

Ncs-architecture.png

The area of networked control systems has emerged in the last several years and seeks to combine some of the insights from computer science and control to allow analysis and design of systems that consist of distributed computation connected together across a network. One example of the architecture for such a system is shown to the right. We are working on a number of areas related to developing fundamental theory that can be applied across a range of networked control systems:

  • Automatic synthesis of control protocols: Next generation networked control systems must also be (at least partly) designed for verification, since it will not be possible to analyze systems of this complexity without structure the design to allow verification tools to be applied. The use of "correct by construction" design methods is one path that shows promise for automatically synthesizing control protocols given a set of specifications describing the required behavior. Preliminary results have demonstrated some possible ways to do this for hybrid systems with nonlinear dynamics and event-driven operations, building on tools developed in the model checking community.
  • Distributed resource allocation: networked control systems often required allocation of shared resources using distributed computation and control. We are interested in how to design algorithms that can be used to allocate power, communications and computing resources to maintain overall properties of the system without violating local constraints.
  • Perception-driven planning: A key element of many networked control systems is the availability of relatively large amounts of sensing compared with traditional control systems. This additional sensing requires sophisticated perception algorithms to interpret, model and make predictions about the external environment, and the development of control algorithms that are aware of the limitations of perception and react properly to conflicting, intermittent or missing information about the environment.

Current projects:


Recent papers: