This paper will present a unique concept implemented in a software design that determines near optimal paths between hundreds of randomly connected nodes of interest in a faster time than current near optimal path determining algorithms. The adaptive pyramidal clustering (APC) approach to determining near optimal paths between numerous nodes uses an adaptive neural network along with classical heuristic search techniques. This combination is represented by a nearest neighbor clustering up function (performed by the neural network) and a trickle down pruning function (performed by the heuristic search). The function of the adaptive neural network is a significant reason why the APC algorithm is superior to several well known approaches. The APC algorithm has already been applied to autonomous route planning for unmanned ground vehicles. The intersections represent navigational waypoints that can be selected as source and destination locations. The APC algorithm then determines a near optimal path to navigate between the selected waypoints.

The level of technology in multiple disciplines that would allow the construction of a vehicle capable of driving itself is near. Several years ago, the US Army recognized a deficiency existed in the image processing technology area and started a research and development program to demonstrate machine vision algorithms needed by unmanned ground vehicles. This technology program has spanned over a 3 year period and has evolved from a laboratory demonstration of the image processing algorithms to designing and developing the hardware necessary to road test the algorithms and access performance. This paper will present a part of this technology effort which implemented intersection detection and navigation using neural net image processing as well as GPS and sensor technology on a military vehicle to demonstrate the feasibility of totally autonomous vehicle navigation.

A new method for simultaneous vehicle localization and dynamic map building is described. The method models the noise sources as bounded distributions and can therefore produce bounded estimates for the vehicle and all the target positions. Correlations that arise between vehicle and target estimates when using other techniques, such as the Kalman filter, do not arise, and hence a significant saving in memory and computation is achieved.

In navigation and tracking problems, the identification of an appropriate model of vehicular or target motion is vital to most practical data fusion algorithms. The true system dynamics are rarely known, and approximations are usually employed. Since systems can exhibit strikingly different behaviors, multiple models may be needed to describe each of these behaviors. Current methods either use model switching (a single process model is chosen from the set using a decision rule) or consider the models as a set of competing hypothesis, only one of which is 'correct'. However, these methods fail to exploit the fact that all models are of the same system and that all of them are, to some degree, 'correct'. In this paper we present a new paradigm for fusing information from a set of multiple process models. The predictions from each process model are regarded as observations which are corrupted by correlated noise. By employing the standard Kalman filter equations we combine data from multiple sensors and multiple process models optimally. There are a number of significant practical advantages to this technique. First, the performance of the system always equals or betters that of the best estimator in the set of models being used. Second, the same decision theoretic machinery can be used to select the process models as well as the sensor suites.

This paper discusses the problem of radar to optical scene matching. Because of the different imaging principles, there exists poor similarity between radar images and optical images, and for region features, only exist the relative stable common characteristics: large scale regions. This paper presents a fast and effective matching method to reach the demand of the fast orientation, which makes use of region segmentation technique to extract large scale regions in a radar image, and recognizes the large scale object in it according to the knowledge of the object region in the corresponding optical image to perform coarse location, finally using an additional template matching processing to perform fine location.

Cognitive computing refers to an emerging family of problem-solving methods that mimic the intelligence found in nature. The common goal of these methods is to crack tough problems that have resisted straightforward analytic solutions, such as intractable problems caused by combinatorial explosions. This paper describes the application of a combination of three of these methods, fuzzy logic, artificial neural networks, and genetic algorithms in a unique manner to provide a solution to rapidly develop flight control systems for unmanned aircraft. The environment resulting from the combination of these three methods has been successfully applied or is currently being applied to the flight control system development for four unmanned rotorcraft: a full scale Bell Helicopter UH-1H aerial target, an American Sportcopter Ultrasport 254 single sear ultralight helicopter, a custom developed 45 pound miniature helicopter operated by the Army at NASA Langley Research Center, and an electronic countermeasures decoy developed at the Naval Research Laboratory. Additional investigations have begun using this approach for the development of flight control system for fixed wing aircraft as either an autopilot for manned flight or as a controller for an unmanned vehicle. This paper gives a broad overview and technical description of these projects.

Autonomous unmanned systems require provision for fault detection and recovery. Multiply-redundant schemes typically used in aerospace applications are prohibitively expensive and inappropriate solution for unmanned systems where low cost and small size are critical. Aurora Flight Sciences is developing alternative low-cost, fault-tolerant control (FTC) capabilities, incorporating failure detection and isolation, and control reconfiguring algorithms into aircraft flight control systems. A 'monitoring observer', or failure detection filter, predicts the future aircraft state based on prior control inputs and measurements, and interprets discrepancies between the output of the two systems. The FTC detects and isolates the onset of a sensor or actuator failure in real-time, and automatically reconfigures the control laws to maintain full control authority. This methodology is unique in providing a compact and elegant FTC solution to dynamic systems with nonlinear parameter dependence, such as high-altitude UAVs (unmanned air vehicles) and UUVs (unmanned undersea vehicles), where the dynamic behavior varies strongly with speed (i.e., dynamic pressure) and density. In simulation, the application of the algorithm to actual telemetry data from an in-flight vertical gyro failure, shows the algorithm can easily detect the failure and further demonstrated (in simulation) reconfiguring of the autopilots to successfully accommodate recovery.

The US armed forces will plan and conduct operational maneuvers from the sea and sustained operations ashore. These deployments will require accurate target acquisition and designation beyond line-of-sight, under limited visibility, and during day or night engagements. A laser target designator/rangefinder boresighted to a FLIR or day camera can perform accurate ranging and target acquisition from an unmanned aerial vehicle (UAV) platform. To meet the mission needs of the US Marine Expeditionary Warfighters, the US Army's and Navy's need for precision targeting for Hellfire, Copper Head, and other laser guided ordinance, the UAV Joint Project Office is evaluating NDI laser designator payloads for UAV application. During the upcoming laser designator evaluation on the Hunter UAV, critical laser parameters will be measured: beam dispersion, energy level, pulse width, stability, amplitude, jitter and payload autotrack performance. The purpose of this and future demonstrations/evaluations is to characterize the laser designators/rangefinders performance on various targets at various ranges. These tests will determine the maximum effective range versus target types and UAV stability parameters. Test data will provide valuable information for future laser designator/rangefinder demonstrations on the Tactical UAV, Predator, and Pioneer.

The modular integrated avionics group (MIAG) is a single unit approach to combining position, inertial and baro-altitude/air data sensors to provide optimized navigation, guidance and control performance. Lear Astronics Corporation is currently working within the navigation community to upgrade existing MIAG performance with precise GPS positioning mechanization tightly integrated with inertial, baro and other sensors. Among the immediate benefits are the following: (1) accurate target location in dynamic conditions; (2) autonomous launch and recovery using airborne avionics only; (3) precise flight path guidance; and (4) improved aircraft and payload stability information. This paper will focus on the impact of using the MIAG with its multimode navigation accuracies on the UAV targeting mission. Gimbaled electro-optical sensors mounted on a UAV can be used to determine ground coordinates of a target at the center of the field of view by a series of vector rotation and scaling computations. The accuracy of the computed target coordinates is dependent on knowing the UAV position and the UAV-to-target offset computation. Astronics performed a series of simulations to evaluate the effects that the improved angular and position data available from the MIAG have on target coordinate accuracy.

Autonomous navigation of a small, slow speed, low altitude unmanned aerial vehicle (UAV) have many potential applications. UAVs are generally used for (i) remote sensing the areas which are difficult to approach, (ii) surveillance, (iii) target designation or jamming, (iv) weapon delivery or as a weapon by itself, etc. Another potential application would be to use them as cost-effective loitering vehicles near the potential enemy sites, creating nuisance value. In most applications, the solution for autonomous navigation is to install inertial navigation systems (INS) on board the flight vehicle and regularly update the INS as often and as accurately as possible. In this paper, different INS updating techniques are briefly mentioned with their advantages and drawbacks, and then a multi-mode image based navigation is proposed. Using several body mounted focal-plane-array imaging sensors, a bigger image is obtained to get sufficient features for matching. The emphasis in this paper is to get vehicle's speed, direction/attitude, and 'running fixes' by using very reliable 'area correlation' tracking. A combination of feature based scene matching along with area correlation is proposed for updating INS. The effort in this paper is to bring out conceptual ideas of image based navigation to make an UAV to perform better and at the same time cost effective.

The unmanned ground vehicle (UGV) demonstration C was completed in July of 1995. This was the third of four planned UGV/Demo II demonstrations. Demonstration C highlighted multivehicle premission planning, mission execution monitoring, multivehicle mobility cooperation, target detection from moving and stationary platforms, obstacle avoidance, obstacle map sharing, stealthy movement, autonomous turnaround, formation control/zone security, cooperative reconnaissance, surveillance, and target acquisition, and hill cresting. This demonstration was the first to have two autonomous vehicles working cooperatively while performing a militarily relevant mission. This paper begins with a background of the UGV program and then focuses on Demo C. The paper finishes with an overview of the Demo II missions.

The advantages of an unmanned ground vehicle (UGV) team include application of UGVs to a wider range of missions, reduced operator workload, and more efficient use of communications resources. Several mission applications for multiple UGV systems are described. Many single-UGV missions can be performed more quickly with a larger group of vehicles partitioning the workload. Certain missions, however, are possible only with a team of UGVs, or are greatly enhanced by cooperation among the vehicles. The four-vehicle surrogate semiautonomous vehicle system developed under the UGV/Demo II program is reviewed, and its capabilities for multi-vehicle operations are described. This system implements unmanned mobility, reconnaissance and surveillance, tactical communications, and mission planning and monitoring. The four semi- autonomous vehicles may work independently or as a team, controlled and monitored by a single operator. Ongoing development efforts for Demo II are described and longer-term directions for multiple UGV systems are presented.

A computer controlled HMMWV automatically retraces a previously recorded path using navigation data, a process we have termed retrotraverse. A Kalman filter combines the output of two navigation systems, an inertial dead reckoning systems and a differential GPS both with and without carrier phase detection. During retrotraverse, the mobility controller uses a velocity controller and pure pursuit steering. Obstacles such as another vehicle can be detected with a laser range imaging device.

In 1994/95, the Project Mustang unmanned ground vehicle (UGV) was used in two field training exercises to introduce soldiers at Ft. Hood, Texas, to the capabilities of developmental UGVs. The UGV used in these exercises was a high-mobility multi-purpose wheeled vehicle (HMMWV) platform, mounted with a mission payload to perform reconnaissance, surveillance, and target acquisition (RSTA), and a robotic vehicle driving package (RVDP) to perform path retrace (called retrotraverse) and teleoperation. The mission scenario at Ft. Hood had a soldier manually deploy the UGV to the mission site and set up the RSTA mission payload. Retrotraverse was then used to assist in retrieval of the UGV at the conclusion of the mission. This paper describes the design and application of the RVDP, used to control the driving of the HMMWV, from initial development in 1990/92 for Demo I to Project Mustang in 1994/95.

As modern unmanned ground vehicles (UGCs) are developed for military field use, the importance of highly reliable and adaptable control processes is becoming evident. A new addition to the research and development UGV community is the technology test bed (TTB), which differs from its predecessors in its unique approach to vehicle control. The TTB vehicle control has three distinct nested control processes. At the outer level is the real-time loop in which the operator sends desired commands over a radio frequency or fiber-optic link to the remote vehicle. The TTB performs the commanded function without operator observable delay making the teleoperated driving and reconnaissance activities intuitive and easy to learn. The middle control loop is the system processor command filtering and distribution loop. In this loop, the incoming operator commands are processed through ;smart driving' algorithms to prevent obvious operator error from endangering the vehicle. Once filtered, the modified commands are distributed along a dual redundant MIL-1553B bus to any of the 10 remote terminals (RTs) on the mobile base unit. Each RT is a self- contained microcontroller capable of performing closed loop control using both fuzzy logic and classical algorithms. The inner control layer is comprised of the closed loops between the respective actuators or sensors and the appropriate RTs. By separating the vehicle's functions into small control processes, each with a local dedicated controller, the vehicle offers a truly distributed control approach. The clear advantages of using the nested distributed control approach include a high level of reliability, significant reduction of unique spare parts, allowance for future expansion,a nd simplicity in the integration of new devices.

Wright Laboratory has been tasked by the Naval Explosive Ordnance Technical Division to develop robotic platforms to perform characterization of areas set aside for ordnance testing. These areas require the identification and removal of the unexploded ordnance before they can be utilized for safe, productive use. The characterization task is performed by autonomously sweeping a designated area with the autonomous tow vehicle (ATV). The ATV tows the multiple sensor platform containing a magnetometer array and a ground penetrating radar. The ATV provides the time and position stamp for sensor data. Analysts then review the post survey sensor data to determine ordnance position. The ATV makes use of several advanced technologies. A hybrid navigation and guidance system using an external Kalman filter delivers vehicle position based on information from a real time centimeter level differential global positioning system and a strapped down laser gyro inertial navigation system. A vision-based obstacle avoidance system helps to account for unknown obstacles during survey. Sophisticated path planning algorithms, and an intelligent software architecture for planning and behavior provide a measure of autonomy. A data collection system controls the functions of the various sensors used for the characterization process and manipulates the data stream to conform to an open ASCII data format and stores it to rugged removable hard disks for later analysis.

The location and removal of buried munitions is an important yet hazardous task. Current development is aimed at performing both the ordnance location and removal tasks autonomously. An autonomous survey vehicle (ASV) named the Gator has been developed at the Center for Intelligent Machines and Robotics, under the direction of Wright Laboratory, Tyndall Air Force Base, Florida, and the Navy Explosive Ordnance Disposal Technology Division, Indian Head, Maryland. The primary task of the survey vehicle is to autonomously traverse an off-road site, towing behind it a trailer containing a sensor package capable of characterizing the sub-surface contents. Achieving 00 percent coverage of the site is critical to fully characterizing the site. This paper presents a strategy for planning efficient paths for the survey vehicle that guarantees near-complete coverage of a site. A small library of three in-house developed path planners are reviewed. A strategy is also presented to keep the trailer on-path and to calculate the percent of coverage of a site with a resolution of 0.01 m². All of the algorithms discussed in this paper were initially developed in simulation on a Silicon Graphics computer and subsequently implemented on the survey vehicle.

This paper presents a system for performing real-time vehicular self-location through a combination of triangulation of target sightings and low-cost auxiliary sensor information (e.g. accelerometer, compass, etc.). The system primarily relies on the use of three video cameras to monitor a dynamic 1 80° field of view. Machine vision algorithms process the imagery from this field of view searching for targets placed at known locations. Triangulation results are then combined with the past video processing results and auxiliary sensor information to arrive at real-time vehicle location update rates in excess of 10 Hz on a single low-cost conventional CPU. To supply both extended operating range and nighttime operational capabilities, the system also possesses an active illumination mode that utilizes multiple, inexpensive infrared LED's to act as the illuminating source for reflective targets. This paper presents the design methodology used to arrive at the system, explains the overall system concept and process flow, and will briefly discuss actual results of implementing the system on a standard commercial vehicle. Keywords: Machine Vision, Self-Location, Autonomous Vehicles, Infrared Sensing, Position Determination

The fuzzy logic adaptive controller for helicopters (FLAC-H) demonstration is a cooperative effort between the US Army Simulation, Training, and Instrumentation Command (STRICOM), the US Army Aviation and Troop Command, and the US Army Missile Command to demonstrate a low-cost drone control system for both full-scale and sub-scale helicopters. FLAC-H was demonstrated on one of STRICOM's fleet of full-scale rotary-winged target drones. FLAC-H exploits fuzzy logic in its flight control system to provide a robust solution to the control of the helicopter's dynamic, nonlinear system. Straight forward, common sense fuzzy rules governing helicopter flight are processed instead of complex mathematical models. This has resulted in a simplified solution to the complexities of helicopter flight. Incorporation of fuzzy logic reduced the cost of development and should also reduce the cost of maintenance of the system. An adaptive algorithm allows the FLAC-H to 'learn' how to fly the helicopter, enabling the control system to adjust to varying helicopter configurations. The adaptive algorithm, based on genetic algorithms, alters the fuzzy rules and their related sets to improve the performance characteristics of the system. This learning allows FLAC-H to automatically be integrated into a new airframe, reducing the development costs associated with altering a control system for a new or heavily modified aircraft. Successful flight tests of the FLAC-H on a UH-1H target drone were completed in September 1994 at the White Sands Missile Range in New Mexico. This paper discuses the objective of the system, its design, and performance.

Many advances in inertial navigation have been made over the last decade. Small, lightweight inertial measurement units (IMU) have been developed which provide suitable accuracy at a reasonable cost for many unmanned systems. Price has also dropped to previously unheard of levels (under 10,000 dollars for large buys). IMUs can be augmented with global positioning system (GPS) receivers to provide highly accurate and robust navigation capability. GPS receivers have also dropped in size and cost and are becoming an attractive option for coupling with an inertial system. GPS systems alone are vulnerable to jamming and are not a good choice for military applications where jamming is a consideration. Current Army policy is not to use GPS as a mission essential element. The focus of this paper will be on low cost IMUs with GPS and their application to unmanned vehicles. In particular, the program to add guidance to the multiple launch rocket system extended range rocket will be discussed.

As the expanding field of navigation technologies allows more accurate guidance of unmanned vehicles (UV), higher performance, smaller size control actuation systems are required. Because of their low cost and minimum maintenance, electro-mechanical actuators have been used extensively in air, ground, and underwater UV. Electronic controllers currently utilized with EM actuators use discrete components integrated on a printed circuit board.

Real time attitude determination is emerging as a promising new application of global positioning system (GPS) technology. Through the use of a special GPS receiver and multiple antennas, which perform differential carrier phase measurements, attitude can be determined as accurate as one milliradian in real-time for dynamic platforms. However, achievable accuracies are a function of the satellite geometry, and until recently the GPS constellation geometry was insufficient to provide continuous attitude data. This paper addresses the feasibility of integrating a GPS attitude determination system (ADS) on an unmanned aerial vehicle through the utilization of software simulation tools, which allow the user to evaluate ADS performance of a dynamic host platform based on satellite visibility.

The performance of real-time differential GPS navigation are affected by the satellite constellation available, the characteristics of user equipment, the data processing method used, the specifications of the data link used to broadcast the data from the reference station to the mobile platform, and the distance between the reference station and the platform. Each of the above parameters are discussed and related errors are quantified. The effect of user equipment characteristics such as code and carrier phase noise, multipath rejection capability, and number of tracking channels, is discussed. The use of latency is quantified as a function of the data processing method and receiver type used. Various aircraft positioning case studies are used to illustrate the capability of the system under various conditions. Attitude determination using a multi-antenna configuration is described, together with limitations due to fuselage stability and wing flexing. The results of various flight test are described to illustrate the performance levels achievable.