David L. Livingston

Toward an Undergraduate League for RoboCup

This paper outlines ideas for establishing within RoboCup a league geared toward, and limited to, undergraduate students. Veterans of RoboCupJunior are outgrowing the league as they enter college and this has motivated us to develop a league especially for undergraduate students – the ULeague. The design of the league, presented here, is based on a simplied setup of the Small-size league by providing standard Vision and Communication packages.

Stimulated Electronic Transition Concept for an Erasable Optical Memory

A new concept for an erasable optical memory is demonstrated using stimulated electronic transition (SET). Large bandgap semiconductors are suitable materials for the SET medium. We have investigated the properties of MgS:Eu, Sm and SrS:Eu, Sm as possible media for the SET process. Quantum storage is achieved in the form of charges in deep levels in the medium and stimulated radiative recombination is used as the reading process. Unlike magneto-optic (M-O) and phase change (PC) processes, optical writing, reading and erasing are achieved without localized heating. The SET process will have an inherently faster data transfer rate and a higher storage density, and the medium will be more durable than the M-O and PC media. A possible application of the SET process in neural networks is also discussed.

Determination of weights for relaxation recurrent neural networks

Abstract A theorem which establishes the solutions of a given optimization problem as stable points in the state space of single-layer relaxation-type recurrent neural networks is proposed. This theorem establishes the necessary conditions for the neural network to converge to a solution by proposing certain values for the constraint weight parameters of the network. Convergence performance of the discrete Hopfield network with the proposed bounds on constraint weight parameters is tested on a set of constraint satisfaction and optimization problems including the traveling salesman problem, the assignment problem, the weighted matching problem, the N-queens problem and the graph path search problem. Simulation and stability analysis results indicate that the set of solutions becomes a subset of the set of stable points in the state space as a result of the suggested bounds. For the cases of the traveling salesman, assignment and weighted matching problems, two sets are equal leading to convergence to a solution after each relaxation. Convergence to a solution after each relaxation is not guaranteed for the N-queens and the graph path search problems since the solution set is a proper subset of the stable point set. Furthermore the simulation results indicate that the discrete Hopfield network converged to mostly average quality solutions as expected from a gradient-descent search algorithm. In conclusion, the suggested bounds on weight parameters guarantee that the discrete Hopfield network will locate a solution after each relaxation for a class of optimization problems of any size, although the solutions will be average quality rather than optimum.

TickBot: A novel robotic device for controlling tick populations in the natural environment

A semi-autonomous 4-wheeled robot (TickBot) was fitted with a denim cloth treated with an acaricide (permethrin™) and tested for its ability to control ticks in a tick-infested natural environment in Portsmouth, Virginia. The robots sensors detect a magnetic field signal from a guide wire encased in 80m polyethylene tubing, enabling the robot to follow the trails, open areas and other terrain where the tubing was located. To attract ticks to the treated area, CO2 was distributed through the same tubing, fitted with evenly spaced pores and flow control valves, which permitted uniform CO2 distribution. Tests were done to determine the optimum frequency for TickBot to traverse the wire-guided treatment site as well as the duration of operation that could be accomplished on a single battery charge. Prior to treatment, dragging was done to determine the natural abundance of ticks in the test site. Controls were done without CO2 and without permethrin. TickBot proved highly effective in reducing the overall tick densities to nearly zero with the treatment that included both carbon dioxide pretreatment and the permethrin treated cloth. Following a 60min traverse of the treatment areas, adult tick numbers, almost entirely Amblyomma americanum, was reduced to zero within 1h and remained at or near zero for 24h. Treatments without CO2 also showed reduction of ticks to near zero within 1h, but the populations were no different than the control sections at 4h. This study demonstrates the efficacy of TickBot as a tick control device to significantly reduce the risk of tick bites and disease transmission to humans and companion animals visiting a previously tick-infested natural environment. Continued deployment of TickBot for additional days or weeks can assure a relatively tick-safe environment for enjoyment by the public.

AN ADAPTIVE CONSTRAINT SATISFACTION NETWORK

A closed-loop adaptive constraint satisfaction network algorithm is proposed. A Hopfield network is used with an adaptation algorithm to solve syntactic constraint satisfaction problems. The inclusion of the adaptation algorithm in the Hopfield network provides a mechanism to detect and eliminate local minima by redefining the energy landscape. The performance of the proposed algorithm is tested on a directedgraph path search problem and is compared with the performances of other existing deterministic, neural networkbased constraint satisfaction methods.

A compact evolutionary algorithm for integer spiking neural network robot controllers

In order to facilitate online training of a robot controller composed of a spiking neural network, we propose the creation of a method dubbed a ‘compact evolutionary algorithm’. The compact evolutionary algorithm, derived from the compact genetic algorithm, greatly reduces the memory requirements for evolutionary optimization and also obviates the need for floating-point arithmetic capabilities allowing its efficient implementation by microcontrollers. The compact evolutionary algorithm is compared to the traditional evolutionary algorithm for solving three cyclic functions that are of use in a walking robot.

A new method for storing weights in analog neural hardware

An optoelectronic method for nonvolatile storage of analog weights in artificial network hardware is proposed. A pulse of light of selected wavelength is used to store analog information by forcing electrons in a suitably engineered medium into stable energy levels. The information is retrieved by exposing the medium to light of a second wavelength which drops the stored electrons back into their initial energy levels, creating a proportionate number of photons which can be detected using a photodiode. This is suggested as a technique, called quantum storage, which can be used to represent analog weights in neural hardware.<<ETX>>

System/370 capability in a desktop computer

A desktop computer with System/370 capability was produced by enhancing the IBM Personal Computer XT with additional hardware and developing software that provides a compatible interface. The computer, the IBM Personal Computer XT/370, and this software allow users to run most System/370 Conversational Monitor System application programs unaltered in a desktop environment. The evolution of the development and details of the function of the hardware and software are described.

A vacuum-tube guitar amplifier model using a recurrent neural network

Rock and blues guitar players prefer the use of vacuum-tube amplifiers due to the harmonic structures developed when the amplifiers are overdriven. The disadvantages of vacuum tubes compared against solid-state implementations, such as power consumption, reliability, cost, etc., are far outweighed by the desirable sound characteristics of the overdriven vacuum-tube amplifier. There are many approaches to modeling vacuum-tube amplifier behaviors in solid-state implementations. These include a variety of both analog and digital techniques, some of which are judged to be good approximations to the tube sound. In this paper we present early results of experiments in using a neural network to model the distortion produced by an overdriven vacuum-tube amplifier. Our approach is to use artificial neural networks of the recurrent variety, specifically a Nonlinear AutoRegressive eXogenous (NARX) network, to capture the nonlinear, dynamic characteristics of vacuum-tube amplifiers. NARX networks of various sizes have been trained on data sets consisting of samples of both sinusoidal and raw electric guitar signals and the amplified output of those signals applied to a tube-based amplifier driven at various levels of saturation. Models are evaluated using both quantitative (e.g., RMS error) and qualitative (listening tests) assessment methods on data sets that were not used in the network training. Listening tests- considered by us to be the most important evaluation method-at this point in the work, are indicative of the potential for success in the modeling of a vacuum-tube amplifier using a recurrent neural network.

Robot navigation using sensors with fuzzy characteristics

A fuzzy controller for navigating a robot tracking a wire is developed. The robot is part of a tick eradication system that works by attracting ticks using CO2 emitted from a tube and collecting and killing them with an acaricide impregnated cloth. The authors show that using sensors with fuzzy characteristics to track a wire in the CO2 tube produces a fuzzy controller that bypasses the fuzzification step resulting in a simple and effective controller.