Beta 1


Title A Model of Bio-Inspired Hardware : The DNA approach
Author Boesen, Michael Reibel
Supervisor Madsen, Jan (Embedded Systems Engineering, Department of Informatics and Mathematical Modeling, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
Institution Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark
Thesis level Master's thesis
Year 2008
Abstract DNA is the blueprint of all living beings. Inspired by this I sat out on a cause to investigate the concept of self-assembling and self-healing hardware by putting a great emphasis in designing a suitable DNA format for the task. In the report I described, analyzed and discussed two distinct DNA formats by simulating the self-assembling and self-healing of circuits in a Java-based simulator I created for this purpose. The basic platform the simulator simulated was a NoC-based platform in which each node in the NoC was a cell. The contents and functionality of the cell was made customizable, such that it was easy to implement the two very different DNA formats. The first DNA format were based on a circuit graph (also known as a block diagram) in which each DNA string represents a node as well as the arcs of that node. The second DNA format were based on a much more abstract model of the circuit to be implemented - namely the algorithm describing the behavior of the circuit. Based on the work of Ian Page [30] I found a way to represent the algorithm and create the programming structures described by the algorithm in hardware. In this way the algorithm(s) of the circuit can be implemented fast on the reconfigurable NoC platform. Self-healing was also achieved by utilizing the information redundancy in the system as well as the fact that all cells can contain the DNA, because of the compactness of the chosen DNA format. With the use of the simulator it was determined that the second DNA format was the best, therefore I designed and analyzed a model of the cell in hardware, which were based on this DNA format. However, the model was not implemented, due to the time limit of my project. The model showed that if each cell only represents the circuit on a gate-level the overhead would be more than 300x! However, if the cellular NoC platform was changed into a rDPA-type, where each cell implements a function on a multiple bit function-level, the overhead would drop to about 2x.
Series IMM-M.Sc.-2008-07
Fulltext
Original PDF ep08_07_net.pdf (7.59 MB)
Admin Creation date: 2008-02-07    Update date: 2010-10-28    Source: dtu    ID: 210322    Original MXD