James F. Frenzel

Special Purpose Image Convolution with Evolvable Hardware

In this paper, we investigate a unique method of inventing linear edge enhancement operators using evolution and reconfigurable hardware. We show that the technique is motivated by the desire for a totally automated object recognition system. We show that an important step in automating object recognition is to provide flexible means to smooth images, making features more obvious and reducing interference. Next we demonstrate a technique for building an edge enhancement operator using evolutionary methods, implementing and testing each generation using the Xilinx 6200 family FPGA. Finally, we present the results and conclude by mentioning some areas of further investigation.

Power-supply current diagnosis of VLSI circuits

This paper presents a technique based upon the power supply current signature (PSCS) which allows testing of mixed-signal systems, in situ. The PSCS contains important information concerning the operational status of the system; such information can be extracted using approaches based on statistical signal detection theory. The fault-detection performance of these techniques is superior to that achieved through autoregressive modeling of the PSCS. These methods are suitable for production testing of cost-sensitive devices and field testing of mission-critical systems. >

A CMOS neuron for VLSI circuit implementation of pulsed neural networks

This article presents a CMOS implementation of a biologically-inspired neuron. The neuron accommodates multiple excitatory and inhibitory inputs with digital weights and generates a pulse-width modulated output waveform of constant frequency based on the level of activation. The behavior of this implementation is demonstrated and it is shown that combinations of neurons form a complete logic set for realizing Boolean functions.

Measurement of magnetic field using collaborative AUVs

This paper describes an effort to adapt a fleet of autonomous underwater vehicles (AUVs) for the measurement of the magnetic signature of surface ships. Currently, two such vehicles have been upgraded with the necessary navigation and measurement hardware to perform magnetic measurements. Initial testing has been performed at the Naval Acoustic Research Detachment at Lake Pend Oreille, Idaho. Two vehicles have leveraged formation control algorithms originally developed for mine countermeasure missions to operate collaboratively. In the testing area, changes in ambient field were measured to be ∼1,000–2,000 nT. Over one square meter area segments, the standard deviation of total field measurements was below 25 nT. When a steel barge was located in the testing area, localized fields of ∼50,000 nT to ∼62,000 nT were observed.

Low-latency multiple clock domain interfacing without alteration of local clocks

A method for interfacing synchronous blocks of logic with different local clocks is presented which introduces very little latency and avoids metastability. The method does not require stopping or stretching local clocks and enforces correct operation of a bundled data constraint for all but very wide data paths.

Representation and robustness for evolved sorting networks

We describe evolved sorting networks for a Xilinx 6200 rapidly reconfigurable Field Programmable Gate Array (FPGA) and for a simulated environment. Our goal was to evaluate the efficiency and stability of evolved circuits in a changing environment. Not only did we evolve correct sorting networks, but we also examined the representations of evolved individuals for their runtime efficiency and effectiveness. We compared three different hardware representations: tree structured encodings, linear direct encodings, and raw configuration files. We also used three separate fitness functions. We also present an interesting metric for gate-level resilience to faults: bitwise stability. We find evidence that evolution inherently improves bitwise stability, and that tree structures may confer more bitwise stability than linear structured chromosomes.

Field Testing of Moving Short-baseline Navigation for Autonomous Underwater Vehicles using Synchronized Acoustic Messaging

This paper presents the results from field testing of a unique approach to the navigation of a fleet of autonomous underwater vehicles (AUVs) using only onboard sensors and information provided by a moving surface ship. The approach, considered moving short-baseline (MSBL) navigation, uses two transponders mounted on a single surface ship that alternately broadcast acoustic messages containing one of the parameters of the kinematic state of the surface ship. The broadcasts are initiated according to a predefined schedule so that the one-way travel time (OWTT) of the acoustic messages may be used to determine the range to the transponder. Each AUV in the fleet uses the surface ship state measurements and ranges provided by the acoustic messages in two extended Kalman filters (EKFs) for state estimation. The first EKF merges the intermittent surface ship state measurements with a kinematic model to estimate the state of the surface ship. This is necessary because the presented approach uses 13-bit acoustic messages as opposed to the more commonly used 32-byte messages, which allow the full state to be encoded in a single broadcast. The second EKF uses the current surface ship state estimate to properly interpret the acoustic ranges, combining them with a kinematic model to estimate the state of the AUV itself. Numerous MSBL navigation experiments were compared against a more traditional approach using a long-baseline (LBL) array of transponders and OWTT acoustic ranging. The results of all tests were verified by independent LBL measures of position.

Size versus robustness in evolved sorting networks: is bigger better?

We are interested in discovering how and why circuits developed using evolutionary methods tend to be more robust than hand designed ones. To this end, we compare evolved circuits to known, minimal sorting networks. We introduce a new, size-independent metric, called bitwise stability (BS), which measures how well a network performed when subjected to real-world types of errors. In particular, we examine stuck-on-one, stuck-on-zero and passthrough errors, a generalization of a short circuit. Networks were evolved using tree structured and linear encoded chromosomes. We found that evolution improves bitwise stability and that tree structures tend to confer more bitwise stability than linear structured chromosomes. We discuss how the size of a sorting network affects its robustness and our discovery that bigger does not necessarily mean better.

A multi-level DRAM with fast read and low power consumption

In this paper, we present a new, multi-level DRAM design, which can store 3 voltage levels (0, Vcc, and Vcc/2) in a single memory cell. This multi-level DRAM requires no special reference voltage and simplifies design of the peripheral circuits. Coding algorithms may be used to provide binary data immediately after first read, with the second read operation providing a second data word from the same cell; thus, binary data from two logical addresses can be obtained from one physical location. One of the coding algorithms uses an additional coding memory cell for every two data memory cells to provide 4-bits of binary data. A second algorithm uses 3 additional coding cells for every 8 data cells to provide 8-bit binary data for each access; thus every 11 memory cells can provide 16 bits of binary data. Furthermore, the read speed is faster than a conventional DRAM because the first access can complete before the word line reaches Vccp, and because a SRAM differential sense amplifier is used. Finally, storing 3 voltage levels in a single memory cell also reduces average power consumption, since the Vcc/2 voltage level requires less write back current than 0 or Vcc voltage level

Ordering autonomous underwater vehicle inspection locations with a genetic algorithm

This paper describes a genetic algorithm for solving the traveling salesman problem (TSP) for autonomous navigation. The method is applied to autonomous underwater vehicles for efficient path planning during underwater mine inspections, sponsored by the Office of Naval Research. This method is significantly easier to implement and much more extensible to real world variants of TSP, e.g. problems incorporating currents, limited turning radius, limitations in depth changes, etc., than other, more efficient, approaches. A specific case study demonstrates a variation accounting for constant currents. Performance is compared against existing behaviors for path planning implemented in the Mission-Oriented Operating Suite (MOOS). The results show that the genetic algorithm performs significantly better than the approach currently implemented in MOOS and successfully accounts for factors such as currents.