This paper reports on work in developing systems of representation for multi-tiered autonomous agent architectures. Although some have suggested that agents may be able to perform complex tasks using stateless behaviors, we believe this to be strawman. Imagine washing the dishes and dropping a fork into a sudsy sink. Even though the fork cannot be perceived, it should still be washed and not forgotten. We believe agents can be made more efficient or effective if they are allowed to use task dependent information about their environment, which comes not only form current perception, but also from expectations or previous perceptions. We do not propose to have agents develop a compete model of their environment, but rather that their current task will decide what elements of the environment are important and what elements need representation.

This paper will provide a summary of the methodology, metrics, analysis, and trade study efforts for the preliminary design o the Vetronics Processing Architecture (PA) system based on the Demo III Experimental Unmanned Ground Vehicle (XUV) program requirements. We will document and describe both the provided and analytically derived system requirements expressed by the proposal. Our experience based on previous mobility and Reconnaissance, Surveillance, Targeting, Acquisition systems designed and implemented for Demo II Semi-Autonomous Surrogate Vehicle and Mobile Detection, Assessment and Response System will be used to describe lessons learned as applied to the XUV in PA architecture, Single Board Computers, Card Cage Buses, Real-Time and Non Real-Time processor and Card Cage to Card Cage Communications, and Imaging and Radar pre-processors selection and choices. We have selected an initial architecture methodology.

Fetch 2 is a team of four small autonomous robots that are under development for the purpose of 'pickup and carry away' unexploded ordnance removal. Individually or as a team, the robots can be tasked by a remote operator to locally search a list of specific coordinates or to perform a full coverage sweep of the terrain. The control strategy is one of supervised autonomy: the robots operate autonomously but support and submit to a 'human in the loop'. The robots perform autonomous behavioral control within a subsumptive framework. A separate operator control unit coordinates the robots to perform the remediation operation efficiently and with minimal inter-robot interference. The platform will provide a unique testbed to address such issues as cooperation and coordination of a team of robots performing a collaborative mission. One control mechanisms of particular interest is the use of 'virtual sensors' to allow external guidance from other robots, a global map, or a human operator to be transparently incorporated into by each robot.

For several decades at the Department of Energy's Savannah River Site, large underground storage tanks have been used to contain highly radioactive waste. This waste must be now transported out of the tanks to be processed into a more suitable long-term storage medium. In addition, the emptied tanks must be cleaned in adherence to both state and federal requirements before being permanently closed. Unfortunately, transfer of the waste by pump leaves behind several types of waste forms away from pump suction: highly alkaline and radioactive sludge, rock-like solid masses called clinkers, or large, solidified salt formations known as tank heels. These waste forms must be dissolved and moved on the tank bottom to pump locations prior to being removed from the tank.

This paper describes the design, development and planned implementation of a system of mobile robots for use in flow through storage applications. The robots are being designed with on-board embedded controls so that they can perform their tasks as semi-autonomous workers distributed within a centrally controlled network. On the storage input side, boxes will be identified by bar-codes and placed into preassigned flow through bins. On the shipping side, orders will be forwarded to the robots from a central order processing station and boxes will be picked from designated storage bins following proper sequencing to permit direct loading into trucks for shipping. Because of the need to maintain high system availability, a distributed control strategy has been selected. When completed, the system will permit robots to be dynamically reassigned responsibilities if an individual unit fails. On-board health diagnostics and condition monitoring will be used to maintain high reliability of the units.

ROBART III is intended as an advance demonstration platform for non-lethal response measures, extending the concepts of reflexive teleoperation into the realm of coordinated weapons control in law enforcement and urban warfare scenarios. A rich mix of ultrasonic and optical proximity and range sensors facilitates remote operation in unstructured and unexplored buildings with minimal operator supervision. Autonomous navigation and mapping of interior spaces is significantly enhanced by an innovative algorithm which exploits the fact that the majority of man-made structures are characterized by parallel and orthogonal walls. Extremely robust intruder detection and assessment capabilities are achieved through intelligent fusion of a multitude of inputs form various onboard motion sensors. Intruder detection is addressed by a 360-degree staring array of passive-IR motion detectors, augmented by a number of positionable head-mounted sensors. Automatic camera tracking of a moving target is accomplished using a video line digitizer. Non-lethal response systems include a six- barrelled pneumatically-powered Gatling gun, high-powered strobe lights, and three ear-piercing 103-decibel sirens.

This paper proposes a person accompanying robot as a novel human collaborative robot. The person accompanying robot is such legged mobile robot that is possible to follow the person utilizing its vision. towards future aging society, human collaboration and human support are required as novel applications of robots. Such human collaborative robots share the same space with humans. But conventional robots are isolated from humans and lack the capability to observe humans. Study on human observing function of robot is crucial to realize novel robot such as service and pet robot. To collaborate and support humans properly human collaborative robot must have capability to observe and recognize humans. Study on human observing function of robot is crucial to realize novel robot such as service and pet robot. The authors are currently implementing a prototype of the proposed accompanying robot.As a base for the human observing function of the prototype robot, we have realized face tracking utilizing skin color extraction and correlation based tracking. We also develop a method for the robot to pick up human voice clearly and remotely by utilizing microphone arrays. Results of these preliminary study suggest feasibility of the proposed robot.

A current limitation in the real-world use of cooperating mobile robots is the difficulty in determining the proper team composition for a given robotic application. Present technology restricts the design and implementation of cooperative robot teams to the expertise of a robotics researcher, who has to develop robot teams on an application-specific basis. The objective of our research is to reduce the complexity of cooperative robotic systems through the development of a methodology that enables the automated synthesis of cooperative robot teams. We propose an approach to this problem that uses a combination of the theories of sensori-computational systems and information invariants, building on the earlier work of Donald, Rus, et al. We describe the notion of defining equivalence classes that serve as fundamental building blocks of more complex cooperative mobile robot behaviors. We postulate a methodology for framing mission requirements in terms of the goals and constraints of the problem, incorporating issues such as multi-robot interference, communication, control strategy, robot complexity, and so forth, into the mechanism. Our initial work restricts the robot application and design space to three multi-robot application domains we have previously studied and implemented: keeping formation, 'mock' hazardous waste cleanup, and cooperative observation. This paper present the foundational ideas upon which our approach to cooperative team design is based.

Research on the coordination of multiple mobile robots has to address three main problems: (i) how to appropriately divide the functionality of the system into multiple roots, (ii) how to manage the dynamic configuration of the system, and (iii) how to realize cooperation behavior. This paper will concentrate on the third aspect. More specifically, the aim of our research is to develop a team of coordinating mobile robots via effective communication for real work applications. We will describe the methodology to achieve cooperative behavior, the experimental mobile robots developed, and potential application areas. The developed system is demonstrated by two examples such as flocking and shared experience learning.

Autonomous mobile robots require wide angle vision for navigation and threat detection and analysis, best served with full panoramic vision. The panoramic optical element is a unique inexpensive first surface reflective aspheric convex cone. This cone can be sized and configured for any vertical FOV desired. The cone acts as a negative optical element generating a panoramic virtual image. When this virtual image is viewed through a standard camera lens it produces at the lenses focal pane a panoramic toroidal image with a translational linearity of > 99 percent. One of three image transducers can be used to convert the toroidal panoramic image to a video signal. Raster scanned CCDs, radially scanned Vidicons and linear CCD arrays on a mechanically rotated state, each have their own particular advantage. Field object distances can be determined in two ways. If the robot is moving the range can be calculated by the size change of a field object versus the distance traversed in a specific time interval. By vertically displacing the panoramic camera by several inches a quasibinocular system is created and the range determined by simple math. Ranging thus produces the third dimension.

The localization and identification of antarctic meteorites is a task of great scientific interest and with implications to planetary exploration. Autonomous search for antarctic meteorites presents a profound technical challenge. Ground Penetrating Radar (GPR) holds the prospect to safeguard antarctic robot from terrain dangers and detect subsurface objects. In January 1998, we validated a 500 MHz GPR sensor as part of a field robotic technology demonstration at Patriot Hills, Antarctica. We deployed the sensor from a sled and integrate with position and attitude instruments to perform field measurements. Data was acquired under different conditions and in multiple locations. The radar detected hidden crevasses from 50 cm. distance, thus showing its merit as a rover safeguarding device. It also localized 5 cm. rocks ins now and ice. Moreover, the radar data was used to characterize snow/ice/bedrock stratigraphy. GPR position measurements enabled ground truth and mapping of the location of hazards and interesting subsurface objects and features.

The state of the art in avoiding obstacles using only vision - not sonar or laser rangefinders - is roughly half an hour between collisions. After reviewing the design and failure modes of several current systems, we compare psychology's understanding of perception to current computer/robot perception. There are fundamental differences - which lead to fundamental limitations with current computer perception. The key difference is that robot software is built out of 'black boxes', which have very restricted interactions with each other. In contrast, the human perceptual system is much more integrated. The claim is that a robot that performs any significant task, and does it as well as a person, cannot be created out of 'black boxes.' In fact, it would probably be too interconnected to be designed by hand - instead, tools will be needed to create such designs. To illustrate this idea, we propose to create a visual depth cues at each pixel, as well as depth cues from neighboring pixels and previous depth estimates. Genetic Programming is used to combine these into a new depth estimate. The system learns by predicting both sonar readings and the next image. The design of the system is described, and design decisions are rationalized.

Accurate motion estimation and reliable maneuver prediction enable an automated car to react quickly and correctly to the rapid maneuvers of the other vehicles, and so allow safe and efficient navigation. In this paper, we present a car tracking system which provides motion estimation, maneuver prediction and detection of the tracked car. The three strategies employed - adaptive motion modeling, adaptive data sampling, and adaptive model switching probabilities - result in an adaptive interacting multiple model algorithm (AIMM). The experimental results on simulated and real data demonstrate that our tracking system is reliable, flexible, and robust. The adaptive tracking makes the system intelligent and useful in various autonomous driving tasks.

Hydraulic excavators are large, powerful machines which are often operated in high-production settings. Successful automation of excavators for mass excavation tasks require safeguarding algorithms which do not negatively impact productivity. We present a two-level sensor-based safeguarding approach which utilizes obstacle detection to prevent collisions and motion detection to halt operation when unanticipated vehicles or people approach the excavator.

Upcoming missions to Mars and the mon call for highly autonomous robots with capability to perform intra-site exploration, reason about their scientific finds, and perform comprehensive on-board analysis of data collected. An ideal case for testing such technologies and robot capabilities is the robotic search for Antarctic meteorites. The successful identification and classification of meteorites depends on sensing modalities and intelligent evaluation of acquired data. Data from color imagery and spectroscopic measurements are used to identify terrestrial rocks and distinguish them from meteorites. However, because of the large number of rocks and the high cost and delay of using some of the sensors, it is necessary to eliminate as many meteorite candidates as possible using cheap long range sensors, such as color cameras. More resource consuming sensor will be held in reserve for the more promising samples only. Bayes networks are used as the formalism for incrementally combing data from multiple sources in a statistically rigorous manner. Furthermore, they can be used to infer the utility of further sensor readings given currently known data. This information, along with cost estimates, in necessary for the sensing system to rationally schedule further sensor reading sand deployments. This paper address issues associated with sensor selection and implementation of an architecture for automatic identification of rocks and meteorites from a mobile robot.

A mobile robot can identify its own position relative to a global environment model using triangulation based on measuring angular separation between three landmarks in the environment. Multiple views from different locations of a smaller set of landmarks can also be used though. Alternatively the current position estimates can be updated using heading and distance measurements to a single landmark. Using these different strategies 8 position estimating techniques have been designed, analyzed and compared. These are based on viewing 1, 2 and 3 landmarks from one or two different viewpoints. It is shown that these procedures may be very sensitive to noise depending on the spatial landmark configuration, and relative position between robot and landmarks. A general analysis is presented which permits prediction of the uncertainty in the triangulated position. The uncertainty measure can be used to determine which of the light different techniques is the most suitable in specific situations. The entire analysis is based on a basic statistical approach, and verified experimentally. In addition to the evaluation of the individual techniques, an algorithm is presented for automatic selection of optimal landmarks. This algorithm enables a robot to continuously estimate its current position from the set of landmarks which provides the most stable solution. It is demonstrated that using this algorithm can result in more than one order of magnitude reduction in position uncertainty.

With the success of Mars Pathfinder's Sojourner rover, a new era of planetary exploration has opened, with demand for highly capable mobile robots. These robots must be able to traverse long distances over rough, unknown terrain autonomously, under severe resource constraints. Based on the authors' firsthand experience with the Mars Pathfinder mission, this paper reviews issues which are critical for successful autonomous navigation of planetary rovers. No currently proposed methodology addresses all of these issues. We next report on the 'Wedgebug' algorithm, which is applicable to planetary rover navigation in SE. The Wedgebug algorithm is complete, correct, requires minimal memory for storage of its worked model, and uses only on-board sensors, which are guided by the algorithm to efficiently senses only the data needed for motion planning. The implementation of a version of Wedgebug on the Rocky7 Mars Rover prototype at the Jet Propulsion Laboratory is described, and experimental results from operation in simulated martian terrain are presented.

Sensor based exploration is a task which enables a robot to explore and map an unknown environment, using sensor information. The map used in this paper is the generalized Voronoi graph (GVG). The robot explores an unknown environment using an already developed incremental construction procedure to generate the GVG using sensor information. This paper presents some initial results which uses the GVG for robot localization, while mitigating the need to update encoder values. Experimental result verify the described work.

This paper describes a hierarchial system for dynamic planning and scheduling among a field of moving obstacles. The solution is directly applicable to many autonomous scenarios, such as movement on a manufacturing floor, navigation of surface/subsurface ships or coordination of air traffic. The overall system is discussed in terms of global/local perception and generative planning/reactive control. The development and implementation of the generative planner portion of the existing system are described in detail. In terms of real-time system performance, the multi-dimensional search problem is made tractable by applying two dependent 2D techniques in series. First, the shortest distance, obstacle-avoiding path to the goal is calculated, then for that path, the time to the goal is optimized. In this space-time representation of the domain, time requirements can be imposed upon the goal(s) and therefore scheduling can also be accomplished.

During June and July of 1997, a mobile robot named Nomad traversed 223km in the Atacama Desert of southern Chile via transcontinental teleoperation. This unprecedented accomplishment is primarily attributed to Nomad's innovative locomotion design which features four-wheel/all-wheel drive locomotion, a reconfigurable chassis, electronically coordinated steering, pivot-arm suspension, and body motion averaging. Nomad's locomotion was configured through systematic analysis and simulations of the robot's predicted performance in a variety of terrain negotiation scenarios. Experimental work with a single wheel apparatus was sued to determine the effect of repeated traffic and tread pattern on power draw. Field test before and during the Atacama traverse demonstrated Nomad's substantial terrainability and autonomous navigation capabilities, and validated theoretical performance projections made during its geometric configuration. Most recently, the augmentation of the internal monitoring system with a variety of sensors has enabled a much more comprehensive characterization of Nomad's terrain performance. Because of Nomad's unique steering design a comparison of skid and explicit steering was performed by monitoring wheel torque and power during steady state turns. This paper summarizes the process and metrics of Nomad's mobility configuration, and reports on experimental data gathered during locomotion testing.

This paper will summarize the Autonomous Mobility system for the Demo III program. The autonomous mobility system involves issues in algorithms, sensors, and processing architectures. We will describe some history, and general philosophies that guided us in the direction of the design described in this paper.

One of the principal roles of the Demo III Experimental Unmanned Ground Vehicle will be as a forward scout performing Reconnaissance, Surveillance, and Target Acquisition (RSTA) operations. This paper will present the elements of the preliminary design process for satisfying the rigorous Demo III ATR requirements, including military vehicle deductibility at a maximum range of 6 km and dismounted soldier detection at 2km. The constituent design issues include sensor selection, sensor suite mounting and stabilization, processing architecture, and algorithm selection. In the context of this selection and design process the lessons learned from previous Unmanned Ground Vehicle RSTA efforts will be introduced and the contractual and subsystem derived requirements will be presented as well as the interface issues for the RSTA subsystem in conjunction with the navigation, mission execution, and communication subsystems.

The robustness of autonomous robotic systems to unanticipated circumstances is typically insufficient for use in the field. The many skills of a human user often fill this gap in robotic capability. To incorporate the human into the system, a useful interaction between man and machine must exist. This interaction should enable useful communication to be exchanged in a natural way between human and robot on a variety of levels. This paper describes the current human-robot interaction of the Stanford HUMMINGBIRD autonomous helicopter. In particular, the paper discuses the elements of the system that enable multiple levels of communication. An intelligent system agent manages the different inputs given to the helicopter. An advanced user interface gives the user and helicopter a method for exchanging useful information. Using this human-robot interaction, the HUMMINGBIRD has carried out various autonomous search, tracking, and retrieval missions.

An approach is presented for the evolutionary development of supervised autonomous navigation capabilities for small 'backpackable' ground robots, in the context of a DARPA- sponsored program to provide robotic support to small units of dismounted warfighters. This development approach relies on the implementation of a baseline visual serving navigation capability, including tools to support operator oversight and override, which is then enhanced with semantically referenced commands and a mission scripting structure. As current and future machine perception techniques are able to automatically designate visual serving goal points, this approach should provide a natural evolutionary pathway to higher levels of autonomous operation and reduced requirements for operator intervention.

We consider a mobile robot that attempts to accomplish a task by reaching a given goal, and interacts with its environment through a finite set of actions and observations. The interaction between robot and environment is modeled by Partially Observable Markov Decision Processes (POMDP). The robot takes its decisions in presence of uncertainty about the current state, by maximizing its reward gained during interactions with the environment. It is able to self-locate into the environment by collecting actions and perception histories during the navigation. To make the state estimation more reliable, we introduce an additional information in the model without adding new states and without discretizing the considered measures. Thus, we associate to the state transition probabilities also a continuous metric given through the mean and the variance of some significant sensor measurements suitable to be kept under continuous form, such as odometric measurements, showing that also such unreliable data can supply a great deal of information to the robot. The overall control system of the robot is structured as a two-levels layered architecture, where the low level implements several collision avoidance algorithms, while the upper level takes care of the navigation problem. In this paper, we concentrate on how to use POMDP models at the upper level.

A new design method for non-time based tracking controller is presented. The key to the non-time based control method is the introduction of a suitable action or motion reference variable other than time, which is directly related to the desired and measurable system output.It enables the construction of control system with integrated planning capability, in which planning becomes real-time closed-loop process. The new design method converts a controller designed by traditional time-based approach to a non-time based controller using action reference. It significantly simplifies the design procedure. The design method is exemplified by a unmanned vehicle tracking control problem. The design procedure and simulation results demonstrate the advantages of proposed method.

The new tendency in mobile robots is to crete non-Cartesian system based on reactions to their environment. This emerging technology is known as Evolutionary Robotics, which is combined with the Biorobotic field. This new approach brings cost-effective solutions, flexibility, robustness, and dynamism into the design of mobile robots. It also provides fast reactions to the sensory inputs, and new interpretation of the environment or surroundings of the mobile robot. The Subsumption Architecture (SA) and the action selection dynamics developed by Brooks and Maes, respectively, have successfully obtained autonomous mobile robots initiating this new trend of the Evolutionary Robotics. Their design keeps the mobile robot control simple. This work present a biologically inspired modification of these schemes. The hippocampal-CA3-based neural network developed by Williams Levy is used to implement the SA, while the action selection dynamics emerge from iterations of the levels of competence implemented with the HCA3. This replacement by the HCA3 results in a closer biological model than the SA, combining the Behavior-based intelligence theory with neuroscience. The design is kept simple, and it is implemented in the Khepera Miniature Mobile Robot. The used control scheme obtains an autonomous mobile robot that can be used to execute a mail delivery system and surveillance task inside a building floor.

This paper presents a universal adaptive (lambda) -tracking control algorithm for wheeled mobile robots moving on the plane in the 3D space. The introduce control algorithm maybe considered as a dynamic version of the PD controller requiring only a knowledge of the robot kinematics. This controller preserves the convergence of the position tracking error of mobile robots to the ball of radius (lambda) > 0, where (lambda) is arbitrary but prespecified. Theoretical considerations are illustrated with simulations.

A navigation system for an autonomous mobile robot is described, which is basically probabilistic. The sensing of the environment is made through a rangefinder, each measurement updating an occupancy grid, which is the internal representation of the robot's environment. The vehicle uses only local information for path planning and obstacle avoidance, searching for the path of minimum cost to the goal. The robot's position and orientation estimates are corrected after each scan of the environment by making a least square fit between the internal map and the set of observed points. Results of computer simulations are shown.

This paper presents a new ultrasonic sensor, intended to be used by autonomous mobile vehicles for ranging, localization and mapping of indoor environments. A novel coding method guarantees proper work of the sensor even if multiple vehicles move in the same room or with disturbing ultrasonic noise sources. The sonar sensor system consists of one emitter and three receivers and can detect 3D targets in real-time. Compensating the effects of temperature and humidity, the system produces sub-millimeter range and sub- degree bearing accuracies up to a range of 3 meters. The coding method was developed to maximize the number of different coding signals as well as the noise suppression and to minimize the signal processing overhead. This coding method and the sensor configuration, hardware and signal- processing are described and the experimental results are presented.

Recently, a proactive crash mitigation system is proposed to enhance the crash avoidance and survivability of the Intelligent Vehicles. Accurate object detection and recognition system is a prerequisite for a proactive crash mitigation system, as system component deployment algorithms rely on accurate hazard detection, recognition, and tracking information. In this paper, we present a vision-based approach to detect and recognize vehicles and traffic signs, obtain their information, and track multiple objects by using a sequence of color images taken from a moving vehicle. The entire system consist of two sub-systems, the vehicle detection and recognition sub-system and traffic sign detection and recognition sub-system. Both of the sub- systems consist of four models: object detection model, object recognition model, object information model, and object tracking model. In order to detect potential objects on the road, several features of the objects are investigated, which include symmetrical shape and aspect ratio of a vehicle and color and shape information of the signs. A two-layer neural network is trained to recognize different types of vehicles and a parameterized traffic sign model is established in the process of recognizing a sign. Tracking is accomplished by combining the analysis of single image frame with the analysis of consecutive image frames. The analysis of the single image frame is performed every ten full-size images. The information model will obtain the information related to the object, such as time to collision for the object vehicle and relative distance from the traffic sings. Experimental results demonstrated a robust and accurate system in real time object detection and recognition over thousands of image frames.

Several Intelligent Vehicle applications require knowledge of the presence of objects in the path of the host vehicle. Current approaches to this in-path decision process do not fully utilize the information available from the forward- looking sensor.An earlier paper described a scene tracking path algorithm in which the observed trajectories of preceding vehicles are used to determine the shape of the road in front of the host, along with the orientation of the host with respect to the road, thereby improving the accuracy of object lane determination. This paper reviews that algorithm and presents simulation-based comparisons of the performances of the conventional and scene tracking approaches.

Passive millimeter-wave detection is advantageous for detection of objects obscured by rain, steam or other aerosols. This coupled together with collision avoidance techniques, based on biologically inspired insect vision models, promises compact low-cost solutions that do not require hardware-intensive image processing. This paper examines a number of possible future directions by identifying trade-offs between different integrated antenna strategies. Signal processing issues are also briefly discussed.

Transmission and storage of traffic video are important to traffic video, algorithms that will take advantage of these properties to ease the video compression should be explored. In this paper, we propose a video compression algorithm that takes advantage of still background of traffic video. By applying the wavelet transform to the compression algorithm, the reconstructed video shows no blocking artifacts as in the discrete cosine transform based approach. A wavelet- based de-blurring preprocessing can also be merged into the video compression algorithm for better reconstructed results if the original video was blurred or foggy.

Non-destructive testing for cracks, fissures, fatigue stress, and corrosion has been demonstrated using eddy- current induced magnetic fields measurable by sensors with Faraday magneto-optic properties. A novel class of such sensors has been developed, the MODE sensor, using Fe-Ga thin-films of the general form (R, Bi)3 (M, Fe)5-12 with R equals (Y, Lu, Tm, or other rare earth ions) and M equals Ga or Al. These films are characterized by very high uniaxial anisotropic field, Faraday rotation, absorption coefficient, and MO figure of merit, significantly improving sensitivity over previous thin film compositions. These properties enable their use in highly compact portable or remotely operated devices and requiring either no eddy current or else brief microbursts of electric current rather than lengthy application of steady current in order to induce magnetic fields within observed structures. A portable system for the testing of bridge structural components, fuel tanks, gas cylinders, and other metallic structures has been designed. This apparatus makes use of a compact portable computer into which video output from the MODE sensor unit received. Using a conventional software interface the operator is able to view the same structure in real time and to apply an array of image processing refinement techniques for improving the resolution of the image. Images may be stored as a constant video stream or as a set of individual snapshots. Additional features that enhance the utility of the system for mobile inspection tasks are discussed. These include the incorporation of a pattern recognition training algorithm and library for operator-enhanced identification of structural defects and condition assessments, as well as the broadcast of image and location data via wireless link to a central server for distribution to consulting engineers and for access of Microstation-type CAD files via a web browser interface.

A miniaturized optical fiber magnetometer has been designed, assembled and evaluated for the detection and classification of vehicles. The sensor element consists of a Fabry-Perot cavity formed between the parallel ends of a high quantity cylindrical metallic glass ribbon and a single mode optical fiber, both held int a hollow tube with an inner diameter several microns larger than the ribbon and the fiber. This sensor head is then potted in a small-diameter, rugged, nonmagnetic housing to allow handling and installation into a protected section of parking lot or other environment to enable vehicle signature detection and analysis. The minimum detectable magnetic field is on the order of 100 nT at dc. The sensor has been used to evaluate the potential detection and analysis of vehicle signatures. When material in a vehicle passes nearby the sensor element, it perturbs the magnetic field and produces a complex output signal dependent upon the shape of the ferrous material in the vehicle, its distance and orientation with respect to the sensor element, and the sped of the vehicle. We have also considered the use of wavelet methods to allow the processing of such data, because it allows variations in differential phasing corresponding to varying vehicles speeds.

The I-Net intelligent embedded sensor architecture enables the reconfigurable construction of wide-area remote sensing and data collection networks employing diverse processing and data acquisition modules communicating over thin- server/thin-client protocols. Adaptive initially for operation using mobile remotely-piloted vehicle platforms such as small helicopter robots such as the Hornet and Ascend-I, the I-Net architecture lends itself to a critical problem in the management of both spontaneous and planned traffic congestion and rerouting over major interstate thoroughfares such as the I-95 Corridor. Pre-programmed flight plans and ad hoc operator-assisted navigation of the lightweight helicopter, using an auto-pilot and gyroscopic stabilization augmentation units, allows daytime or nighttime over-the-horizon flights of the unit to collect and transmit real-time video imagery that may be stored or transmitted to other locations. With on-board GPS and ground-based pattern recognition capabilities to augment the standard video collection process, this approach enables traffic management and emergency response teams to plan and assist real-time in the adjustment of traffic flows in high- density or congested areas or during dangerous road conditions such as during ice, snow, and hurricane storms. The I-Net architecture allows for integration of land-based and roadside sensors within a comprehensive automated traffic management system with communications to and form an airborne or other platform to devices in the network other than human-operated desktop computers, thereby allowing more rapid assimilation and response for critical data. Experiments have been conducted using several modified platforms and standard video and still photographic equipment. Current research and development is focused upon modification of the modular instrumentation units in order to accommodate faster loading and reloading of equipment onto the RPV, extension of the I-Net architecture to enable RPV-to-RPV signaling and control, and refinement of safety and emergency mechanisms to handle RPV mechanical failure during flight.

The main objective of this article is to develop techniques of robust control and their applications to communication and transportation networks. Three following problems will be studied in detail: (1) Resource Sharing: These systems belong to a class of hybrid systems where the plant to be controlled is a continuous drift free system, and the controller is implemented as a FSM. (2) Admission Control. Ramp metering control problem can be studied in either distributed or lumped parameter setting. These two settings to an isolated case have been studied by the authors and solutions for these have been obtained. (3) Traffic Routing. Traditionally both kinds of network systems have been treated using static optimization methods for congestion control. The traffic routing problem for both types of networks can be solved either in user-equilibrium setting or system-optimal. In user-equilibrium, the aim of the controller is to obtain equal travel times on alternate routes. In system-optimal, the aim is to obtain minimum total travel time on the entire network.

This paper presents a prototype incident management decision support system developed using Java development tools. The incident duration and delay estimation models and heuristic resource allocation rules used in the development of the prototype are based on the models used in the Wide-Area Incident Management Support System (WAIMSS) software developed at the Virginia Tech Center for Transportation Research. The Java based application present various advantages over the UNIX based WAIMSS due to its use of one of the common Web browsers as the user interface and object oriented nature of the Java language that allows the re-use of applets for developing the prototype. Although the prototype presented in this paper is not a full implementation of WAIMSS and lacks its Geographical Information Systems capabilities, it clearly demonstrates that Java and Internet will the natural choices for the development of similar real-time decision support systems in the future.

With the advent of the Global Position System (GPS), we now have the ability to determine absolute position anywhere on the globe. Although GPS system work well in open environments with no overhead obstructions, they re subject to large unavoidable errors when the reception from some of the satellites is blocked. This occurs frequently in urban environments, such as downtown New York City. GPS systems require at least four satellites visible to maintain a good position 'fix', and tall buildings and tunnels often block several, if not all, of the satellites. Additionally, due to selective availability, where small amounts of error are intentionally introduced, GPS errors can typically range up to 100 ft or more. This paper proposes several methods for improving the position estimation capabilities of a system by incorporating other sensor and data technologies, including Kalman filtered inertial navigation system, rule- based and fuzzy-based senors fusion techniques, and a unique map-matching algorithm.

We have proposed, investigate,d and implemented a general framework for the simulation, verification, and proto-typing of control algorithms for intelligent vehicles and highways. prior to this project the protocols and control algorithms should have been manually verified, translated to a simulation language for simulation, and then modified for the QNX real-time operating system for porting to the vehicle's computer. This manual translation process is error prone at every stage. Our framework performs the translations automatically, and therefore, removes the possibility of the translation errors. The specification of the control algorithms is performed in the SHIFT specification language. An existing verification platform is used to carry out the correctness proofs of the control algorithms. The QNX real-time operating system, which is currently in use at PATH in automated vehicles, is used as the target platform for the generated code.

This paper presents a feedback control design for isolated ramp metering control. This feedback control design, unlike the existing isolated feedback ramp controllers, also takes into account the ramp queue length. Using a nonlinear H(infinity) control design methodology, we formulate the problem in the desired setting to be able to utilize the results of the methodology.

Many biological system exhibit capable, adaptive behavior with a minimal nervous system such as those found in lower invertebrates. Scientists and engineers are studying biological system because these models may have real-world applications. the analog neural controller, herein, is loosely modeled after minimal biological nervous systems. The system consists of the controller and pair of sensor mounted on an actuator. It is implemented with an electrical oscillator network, two IR sensor and a dc motor, used as an actuator for the system. The system tracks an IR target source. The pointing accuracy of this neural network controller is estimated through experimental measurements and a numerical model of the system.

The Mobile Detection Assessment and Response System, Exterior (MDARS-E) provides an automated robotic security capability for storage yards, petroleum tank farms, rail yards, and arsenals. The system includes multiple supervised-autonomous platforms with intrusion detection, barrier assessment, and inventory assessment subsystems commanded from an integrated control station.