Robotics 98

Serpentine robots offer advantages over traditional mobile robots and robot arms because they provide enhanced flexibility and reachability, especially in convoluted environments. These robots are well suited to inspect large space-fairing truss structures such as the future space station and can also be used to inspect the Space Shuttle cargo bay before launch. Serpentine mechanisms offer unique capabilities on Earth to applications such as search and rescue, surface coating, and minimally invasive surgery. The work, described in this paper, will exploit a geometric structure, termed a roadmap, to guide the motions of a serpentine robot in highly convoluted spaces. This approach offers advantages over previous work with serpentine robots because it provides a general mathematical structure that is not mechanism specific, thereby having applications to a large class of problems.

This paper describes the design and analysis of an Output Feedback Sliding Mode Control (OFSMC) architecture for flexible manipulators. The results of this study include the analytical development and experimental verification. The hardware system investigated included, (1) a modular flexible one- and two-link planar manipulators with conventional strain gages strategically located on each link for strain sensing, (2) DC servo motors and incremental encoders for each degree-of-freedom used for rigid motion slewing, and (3) a MATLAB/SIMULINK/dSPACE system for real-time control and data acquisition. A nonlinear observer based on the dynamic model was used to determine system states for the model-based portion of the control law for the flexible single-link. Cubic spline reference trajectories were used to slew the manipulator. A numerical simulation model was matched with the physical parameters with the experimental response to a reference input trajectory. A nonlinear optimization algorithm was used for the two-link setup to determine optimal gains from the calibrated dynamic model. The OFSMC was able to produce a rest-to-rest, residual vibration-free, large angle maneuvers. These OFSMC designs were verified on the Sandia National Laboratories flexible manipulator testbed. These results show reductions in both residual vibration and settling time.

Actuation devices are used for many space applications and there is increasing need to reduce their size, mass, cost and power consumption. To address this need, JPL is developing electroactive polymers (EAP) with emphasis on two EAP categories that induce large bending and longitudinal actuation strains. Comparison between EAP and the widely used transducing actuators shows that, while lagging in force delivering capability, these materials are superior in mass, power consumption and actuation strain levels. This study is concentrating on the development of effective EAPs and the enabling of mechanisms that employ their unique characteristics. Several EAP driven mechanisms that emulate human hand were developed including a gripper, manipulator arm and surface wiper. The manipulator arm is made of a composite rod that is lifted by a longitudinal rope actuator and has an end-effector gripper with bending EAP fingers allowing to grab and hold such objects as rocks. The EAP surface wiper operates like a human finger and can be used to remove dust from windows and solar cells. These EAP driven devices take advantage of the large displacement capability with less concern for their limited force actuation capability.

A joint trajectory for point-to-point motion control of a flexible-link robot with a tip loaded mass is derived in this paper. A feedback control system is designed so as to make the robot joint follow this trajectory which results residual vibration-free response. The design is simple. The finite time duration in which the desired motion of the joint angle is achieved, along with elimination of the residual vibration, can be arbitrarily specified. This offers options for various practical considerations. Experimental results have confirmed the effectiveness of the control approach presented in this paper.

A nonlinear oscillator control scheme is proposed to implement basic behaviors in autonomous mobile robots. The rationale for such a system is discussed, and simulations of obstacle avoidance are used to demonstrate the viability of the method. A learning algorithm is demonstrated that indicates a practical method for determining oscillator parameters is available.

Reliable indoor navigation has been a goal of mobile robotics research in recent years. Probabilistic belief update, occupancy grid methods, and multi-tiered hybrid architectures have all been implemented in the pursuit of autonomous navigation systems that can function in the face of danger and incomplete information. The conventional approach applies AI techniques to the localization problem in order to attain reasonable navigation reliability. In this paper, we describe a simple navigation system that is neither multi-tiered nor AI-based. The system has been implemented on a Nomad 150 mobile robot that has demonstrated statistically significant navigation reliability during working hours in several buildings in the San Francisco Bay Area. In addition to providing navigation benchmarks based on long-term empirical tests, NaviGates demonstrates implementations of robot skills that are essential to robot autonomy: robust dynamic obstacle avoidance, path replanning in case of obstruction, automatic avoidance of dangerous regions such as staircases, and automatic human-guided mapping for a new building, allowing the robot to be introduced to a new office building without the need for time consuming manual mapping. Simplicity itself is an important thesis of this research, as a robot with an extremely simple control algoritm is shown to perform admirably in a variety of environments.

We consider the problem of finding optimal long-distance routes for robots in a terrain with approximate information. The terrain is characterized by the average speed that can be maintained by the robot in different parts of the landscape. Additional specified terrain properties might be, for instance, the level of threat or visibility. The robot is required to find a route that corresponds to an optimum in terms of the terrain conditions. Examples of optimal routes are those with shortest travel time, those with minimal exposure to danger, those with a lowest or highest visibility, etc. A complete solution of the problem is obtained by an algorithm that is based on Huygens' principle of wave propagation. The algorithm has been developed for static terrain with a fixed destination point, but, because of its simplicity, it can be used also when the terrain is dynamic. Possible application areas of the algorithm are path finding for scouting robots in a battlefield, finding optimal routes for aircraft in hostile or dangerous environments, finding optimal routes in street maps, and, generally, finding routes that satisfy some optimality condition in an environment represented by maps.

The objective of this work is to bring a response to the safe paths planning problem. So, a path planning method dealing with localization uncertainties is proposed where both uncertainty in position and in orientation of a nonholonomic mobile robot are managed. The safeness of this method is due to the melting between the planning phase and the navigation phase by use of the same process of localization (the Kalman filter). This process realized the multisensor fusion of data provided by a ring of both telemetric and contact sensors in combination with a dead-reckoning system. Numerous examples of planned paths are presented showing the behavior of the mobile robot according to localization information.

Path planning algorithms determine a path from start to goal locations, generally minimizing a cost such as time, distance, fuel, or other parameters. Often, this cost is estimated by the straight-line distance between two points or configurations. However, for a free flying space robot, a straight line is not necessarily the least costly path, because discrete thrusters generate arc-shaped trajectories defined by the current orbit of the robot. So, a way to consider non-straight paths must be found. This research uses the geometries in Voronoi diagrams to determine and define the path to take for a free flying space robot called AERCam. The current work deals with a two-dimensional world, though the extension to three dimensions is planned. Object-free corridors through which AERCam may safely maneuver are also defined. This work will lead to methods for determining the optimal path.

Animals are born with certain innate (preprogrammed) knowledge about themselves and their environment. These simple "rules of life" help them to survive their harsh environment and evade predators. For example, when baby sea turtles hatch from their eggshells, they instinctively dig themselves out of the sand and head toward the safety of the sea. These instinctive or innate behaviors are also present in other animals. If animals have certain innate rules that help them start their lives off successfully, shouldn't autonomous mobile agents also have some innate rules? The learning algorithms presented in this paper depend on some innate knowledge about the environment and the physical properties of the agent's embodiment. Using simple innate knowledge and expectations, the agent will be able to autonomously configure its motors and its sensors. In short, this paper presents Innate Learning (IL) through the realization of two IL algorithms and demonstrates how IL can be used to simplify the learning process.

Virtualization is a technique for the autonomous, non-contact creation of a virtual environment from an existing, real environment and represents a new approach to remote sensing and for providing telepresence in previously unmodeled environments, e.g., waste disposal and storage areas. The emphasis in this paper is on geometric aspects of site virtualization. The topics of ellipse and quadric parameter estimation, including error measures and alternate parameter normalizations are addressed.

In this paper, we present a Co-operative Multiple Robots Survey System (CMRSS) for traversing an unknown area to achieve complete coverage. Each robot is equipped with obstacle detection and own positioning sensors as well as communication equipment to exchange information with other robots. We will describe how CMRSS organizes the multiple robots based on Contract Net Architecture and the high-level communication protocol designed to achieve co-operation among multiple robots. We also describe the architecture and design concept of the high-level software module of CMRSS called Cooperative Mission Execution System (CMES). We have built a simulation environment to serve as the test-bed and implemented the core modules of CMRSS. Some simulations have been done and results have shown the effectiveness of our approach. Implementation on real robot is being planned.

A technique for the autonomous, non-contact creation of real-time navigable, virtual environments from existing, real environments is discussed. The technique, termed virtualization, uses structured light, together with assumptions restricting the classes of target surfaces, to provide direct estimation of the geometry of 3-D surface patch parameters. Passive camera images are used in conjunction with the deduced geometry to generate surface colors and textures. Color and texture recognition are achieved through a partial inversion of the rendering equation that is used to build the virtual environments. Texture samples are extended to complete texture maps by discrete cosine transform. An example reconstruction is provided.

This paper presents some of the solutions that have been and continue to be developed at the Lawrence Livermore National Laboratory (LLNL) to address the technical challenges associated with operating in nuclear and hostile environments. The nuclear challenge and, in general, most hostile environments present unique circumstances and conditions requiring new approaches and techniques to finding solutions. Approaches and solutions used in controlled or conventional environments become limited when used in the highly volatile nuclear environment. Accomplishing tasks in nuclear and hostile environments presents challenges to the approach in controlling hardware and processing materials. Any approach identified must not radically disturb the surrounding environment. LLNL engineers have been actively involved in finding unique and creative intelligent automation solutions to tame the hostile environment. Significant advances have been made in robotics systems, sensors technology, and material handling algorithms. This paper describes these recent advances in technology development and applications engineering.

A waste retrieval system was prepared and tested by GreyPilgrim LLC, Waterjet Technology, Inc., and NIST on behalf of the Hanford Tanks Initiative. This system was tested for static deflection, vibration response, and control system effectiveness, and was used in a demonstration of waste retrieval. Performance of the manipulator was better than anticipated by the project team, and as of this writing the manipulator is still in service eight months after demonstration. This report summarizes the results of this testing, and briefly highlights near-term development plans for the manipulator.

The Hanford Tanks Initiative (HTI) project contracted with commercial vendor teams to demonstrate retrieval technologies for removing radioactive and hazardous waste from Hanford single-shell tanks located in Richland, Washington. The goal of this testing was to address issues related to tank waste retrieval to minimize the risk, uncertainties, and, ultimately, reduce the overall cost of removing waste from the Hanford tanks. Two of these demonstrations involved the use of long reach manipulators to deploy waste retrieval end effectors. This paper highlights the manipulator-based retrieval systems demonstrations and analyzes the testing issues that need further resolution.

This paper describes the Remote Area Disassembly Vehicle (RADV) System and its operation on the test range. The RADV System goal is to optimize test data collection and reduce safety and environmental hazards during impact area operations. The RADV System consists of two remote control vehicles and one remote control backhoe, each equipped with video cameras, and operated from a control van. This system serves to retrieve, disassemble, analyze, and dispose of explosive devices tested at the Warheads Impact Target (WIT) areas at White Sands Missile Range (WSMR).

The Air Force Research Laboratory (AFRL) in cooperation with Headquarters Air Combat Command and the 99^th Civil Engineering Group, Nevada Test Ranges has developed a robotics platform capable of carrying a suite of munitions disposal and range maintenance tools. The purpose of this paper is to outline the development and testing of the system and advocate continued research into the automation of UXO clearance, range maintenance, remediation and force protection. The All-purpose Remote Transport System (ARTS) consists of a low cost survivable platform capable of remote operations at distances of up to three miles. Two versions of this system have been fielded: one for active range clearance and one for force protection.

A potential problem in the production of cement testing cubes is inconsistency in quality due to skill differences between operators and low repeatability in human performance of identical operations. To eliminate this problem and to enhance productivity, a state-of-the-art robotic workcell system, which utilized a multitasking control strategy and tool changer and sensor technology to automatically produce cement testing cubes, was proposed, designed, and integrated. A comparative analysis of compressive strength values of specimens made by human operators and robots indicated that the specimens made by the robotic workcell system had lower variation than the human made ones. In addition, the system was designed to (1) comply with the American Society of Testing Methods (ASTM) procedure requirements, (2) pass the cement and concrete reference laboratory inspection checklist, and (3) be a fully automated turnkey system. This study not only demonstrates that robotic workcells are flexible enough to be used in cement testing cube production, but also suggests that revision of ASTM procedures to facilitate implementation of high technology in the materials testing and measurement process should be considered.

Efficient and accurate operation of heavy machines with coordinated motion in multiple axes requires advanced computer-based control approaches to take advantage of the hardware performance. In this context the development of appropriate human-machine-interfaces (HMI) that meet the specific demands from both the human operator and the rough environmental conditions is a crucial point. The increasing complexity of such systems demands for extensive assistance in manual operation as well as for extensive monitoring. In this paper we present the development of an adaptable human-machine-interface for the operation and supervision of heavy machines demonstrated by a Cartesian controlled drilling boom. The first two sections introduce the system and the modifications for the control system integration outlined in the third section. The description of the HMI and its extension with respect to a distributed control approach will be followed by an illustration of the monitoring concept and its implementation.

Reports about serious injuries and costly damages caused by excavators hitting underground utilities make frequent headlines in the news media. The accidents sometimes result in the death of one or more persons. Thus, excavation represents a dangerous operation that has to be executed with care. The Construction Automation and Robotics Laboratory (CARL) at North Carolina State University (NCSU) has been searching for an answer to the problem. A prototype system has been developed in CARL using nontraditional tactics. It is called a Buried Utility Detection System (BUDS). BUDS differs from the traditional passive metal detection systems which require the existing utility lines to serve as transmitters. BUDS generates and transmits its own magnetic impact and detects the coupling effect with any buried utility line in its detection range. Most importantly, it is installed directly on the excavating machinery and integrated with its operation. Experiments have been carried out with BUDS both in the laboratory and in the field. These experimental results are promising. The work deliberated in this paper presents an ongoing effort for developing an effective and reliable system that can be attached to any type of construction and utility digging equipment for real time underground utility line detection.

Earth moving is one of the most complex processes in the construction industry. Nevertheless, there is a strong trend towards excavation automation. Recently, more attention has been paid to trajectories of a bucket during excavation processes. The aim of the present paper is to discuss the question of the shortest path between two points of the working zone. The problem is presented as a non-linear programming problem. The objective function presents the length of the path that has to be minimized. Then several equality and inequality constraints are imposed on the problem. First of all, it is assumed that the investigated trajectories are plane curves. Next, traveling between two points, the bucket is filled with the soil to a given volume. This in turn requires knowledge of the free soil surface. The second set of equality constraints are equations joining lengths of the excavator arms and their angles of rotations. The latter are then related to the lengths of hydraulic actuators. Inequality constraints cover relations restricting rotations of the bucket. Moreover, the depth of the tip penetration in soil is restricted according to the soil properties. Finally, inequalities deal also with boundary conditions at both extreme points of the trajectory. They take under consideration possible penetration of the tip at one of the end of the curve, and readiness to keep the soil in the bucket at the second end. The paper is illustrated with examples of earth removal from deep excavations.

The Construction Automation and Robotics Research Laboratory of USAF Wright Laboratory Air Base Technology Branch (WL/FIVC) is currently developing technologies and robotic systems to be used in hazardous operations. Each system requires a robotic vehicle that can be controlled remotely or can operate autonomously. In the simplest case, teleoperation, the vehicle requires an on-board control computer linked to a joystick box via RF modems. For autonomous operation, other components must be added to the vehicle and to the Operator Control Unit (OCU). Autonomous operation requires the addition of the Mission Specific Subsystem (MSS) on the vehicle and a display computer on the OCU. The Vehicle Control Unit (VCU), Autonomous Interface Unit (AIU) and the MSS components have been designed to be as generic as possible with respect to various vehicles. On previous robotic vehicles, a unique computer system and software has been developed on a case-by-case basis. Our current modular system is comprised of a control computer, software, and external sensors that can be applied on a generic basis from one vehicle platform to the next reducing duplication of effort. This Robotic Control System (RCS) has been developed using the DeviceNet Controller Area Network (CAN bus). The VCU and AIU use rugged embedded PC/104 hardware and the MSS is based on VME multi-processor technology. The DeviceNet network is a serial bus with a data rate of 500 Kbps and eliminates discrete point-to-point wiring to connect sensors or actuators to the control computers. The vehicle control and sensor interface software resides on solid state disks in the VCU and AIU, which automatically boots upon power up. The software, upon initialization reads and parses an ASCII text vehicle configuration file. This file describes the vehicle and the associated DeviceNet nodes. When a new vehicle is to be made robotic, this configuration file is virtually the only software component on the VCU and AIU that needs to be modified.

Essential part of intelligent sensory systems for robotics and man-machine interface is based on the scanning of six degrees-of-freedom information about three axial shiftings and three angular displacements. Simple construction of six-DOF force-torque sensor, hand controller and chaser for autonomous vehicles is based on the modular sensory brick-box system, enabling the customization. The system for the scanning of six-DOF information is based on the CCD element and four light beams or planes from the laser diodes, creating the shape of the pyramid.

A highly maneuverable underwater vehicle that can be operated remotely with computer assistance was designed, fabricated and tested by researchers at the University of Florida. The vehicle incorporates many off-the-shelf components in order to minimize cost. The body was constructed using a foam core with a fiberglass/carbon fiber outer shell. Energy was provided by an Exidel 12V gel cell wheelchair battery. Four trolling motors, two oriented horizontally and two vertically, provide forward/backward thrust, turning, and ascend/descend movement. Attitude and heading are furnished by Precision Navigation's TCM2 digital compass. Vertical position in the water is sensed with a pressure gauge. Two valves are used to fill a buoyancy compensator with air from ballast tanks, or to operate an air-actuated valve to release the air and fill the buoyancy compensator with water. A 68HC11 microcontroller from Novasoft is used to read in data from the sensors, and provide controlling signals to the motors. Approximately 0.9144 m (3 feet) wide by 0.9144 m (3 feet) long by 0.6096 m (2 feet) high, the sub is highly maneuverable due to its small size and tight turn radius. Weighing about 40.82 kg (90 pounds) out of the water, it is close to neutrally buoyant in the water. Since buoyant forces are near the top of the sub, while the heavy weight is located at the bottom, it is also inherently stable. The paper will detail the vehicle design and the resulting performance capabilities.

As part of various research and development activities in the area of mobile planetary robots small vehicles with masses of 10 kg or less are currently receiving high attention. In the frame of future solar system exploratory missions the European Space Agency (ESA) has identified the need to extend the range of scientific measurement from the immediate vicinity of static landing stations to a wider radius of a few tens of meters. This paper highlights the on-going development of a small mobile robot within the framework of ESA's Technological Research and Development Program. The purpose of the vehicle is to deploy instruments or sensor heads, acting as an `extended-robotic arm' rather than covering large distances on various types of terrain, while maximizing the mass allocation for scientific equipment. Applications of the vehicle could be the projected European mission to Mars, called `Mars Express', to be launched in 2003, and future European missions to the Moon as well as potential international missions. The requirements call for a simple, robust device with a net vehicle mass of only 2 kg and with low power consumption. Assuming a model payload, several candidate concepts have been defined, followed by a trade-off and selection of two different configurations - one based on tracked and one based on wheeled locomotion. This paper describes the main requirements, the design and the main test results for the tracked vehicle called `NANOKHOD', while the wheeled vehicle is presented in a complementary paper.

Mobile robots are of high interest for unmaned planetary exploration. The very successful Pathfinder mission to Mars has impressively demonstrated the potential of mobile platforms for planetary exploration. The European Space Agency (ESA) also started to develop concepts for micro-rovers for Mars missions. Within an interdisciplinary group of companies specialized in space applications and research labs new designs of micro-rovers have been investigated. Two concepts, a simple and robust one and an innovative one, have been selected and functional breadboard models of them are currently built. After a discussion of the key issues for robust locomotion the present paper will focus on the design and control of the more innovative solution. It consists of 6 independently driven wheels arranged in two triangles. It therefore allows not only for efficient rolling on flat surfaces but also to step on obstacles. Additionally the center of mass and the instrumentation carrousel is adjustable, allowing to optimally balance the micro-rover in almost every situation. Even after flipping over the robot will always be able to get back on its wheels.

Excavation, and in particular trenching using backhoes, represents a hazardous working environment for humans. Many hazards exist including trench walls which can collapse, heavy objects which can be accidentally dropped into the opening, and buried utilities that are damaged during the operation. The objective of this paper is twofold: (1) present the concept and development of a spatially integrated pipe-laying system, and (2) prove the technical feasibility and effectiveness of such a system under field conditions. A teleoperated device has been developed which would perform the pipe-laying operation. The pipe-manipulator was integrated with a backhoe excavator, control units and a laser based position measurement tool. An innovative technology for remotely controlling large concrete pipe laying was tested. This technology is considered a key component for improving workers' safety and productivity.

This paper describes the sensing portion of a system used to convert an existing 30-ton bridge crane to computer control for automated placement of construction components. The system is being designed to permit either telepresent or fully autonomous assembly of representative parts of buildings and industrial plants, e.g., girders, tanks, and pipes. In order to achieve six degree-of-freedom manipulation of the components, the traditional crane cable and hook have been replaced by TETRA, an inverted cable-operated Stewart platform equipped with various manipulators. The sensing systems are primarily displacement, force, and other state sensors. What makes this application unique is the scale of the robot: the crane's workspace is 40 m long, 23 m wide, and 24 m high — a volume of more than 10,000 cubic meters. This scale, and the need to operate without wires has led to a control system based on wireless packet communication by intelligent modules known as "smart pods." The pods interface to one or several force, displacement, or state sensors. The pods broadcast their data via wireless ethernet to base stations which then communicate with the world via higher-speed networks. The present paper discusses issues relating to the architecture of the sensor array needed to operate this large construction robot and the communications infrastructure needed to supply that information to remote sites in real time.

This paper describes basic concepts to develop the remote controlled construction manipulators system for transmission towers. The purpose of this system is to reduce severe working condition during tower assembling process that involves some risk of fall accident. Remote controlled assembly technology is a key for future large-scale transmission tower construction. Authors have developed a concept of the semiautomatic construction manipulators system for transmission towers.

An autonomous wheeled robot may collide with rocks found on its traverse over rough terrain. The purpose of this investigation is to determine the change in motion of a wheel under vehicle weight resulting from impact with an obstacle, and the impulse imparted at the region of contact. A combination of rolling and slipping motions is considered before and after collision. The wheel is simplified to a disc, and the local terrain to a horizontal plane. It is shown that if the disc is in point contact with the plane, assumes a vertical configuration, and moves at constant rate along a straight path to the obstacle, the motion is valid in a mathematical sense. The boundary condition of point contact, however, leads to the following paradox: Without acceleration, friction does not exist, yet there must be sufficient surface traction to support rolling. In addition, the work done by the vector force of constraint per unit displacement of the disc is zero, hence no change in kinetic energy can occur. The modification to a finite region of deformable surface contact resolves these inconsistencies. Under vehicle weight, a flat contiguous region is formed in which the outermost layers of the wheel are compressed and strain energy is stored. A fixed number of material elements located on the wheel periphery undergo a change in direction from circular paths to nominally horizontal ones. At exit from the region of contact, strain energy is released and the wheel configuration is restored. Rotational rate, mass center velocity, the two components of impulse, and the loss in kinetic energy are found as functions of two slippage parameters and the rotational rate prior to impact. The parameters refer to a propensity to slip first on the terrain, and then on the obstacle.

Terrain adaptive wheeled vehicles offer great promise for locomotion on rugged terrain. In order to properly design terrain adaptive vehicles or control schemes for such vehicles, one needs an effective simulator. Many simulators are based on detailed dynamic analysis. Featherstone's method of articulated inertias makes such dynamic simulators reasonably efficient. However, rover vehicles often operate at low speeds that make the dynamic analysis unnecessary. A simpler approach is to use only velocity and position information to obtain the next state. Simulation on rugged terrain requires the program to account for gravitational and contact forces. An algorithm using potential energy methods is proposed as an alternate rough terrain simulator.

We solve the Motion Planning Problem for nonholonomic systems without drift under a discrete levels constraint, i.e. when it is driven by input that can assume only discrete levels. In particular, we propose an algorithm that steers a drift-free system from a given initial point to a given final point within a finite time interval and we prove that the proposed algorithm provides an exact steering when the system is nilpotentizable. For a general system without drift, i.e., for a drift-free system that is neither nilpotent nor feedback nilpotentizable, the discrete levels control can be computed by an iterative algorithm. Finally, we explore the details of the proposed methods through a specific example.

Modern mechanical systems are increasingly being controlled by digital electronics, yet many mechanical engineering programs have not incorporated this topic into the curriculum. Also, ABET has recently emphasized the importance of design in engineering education. The LEGO Robot Competition is a course offered by the Mechanical Engineering Department which addresses the integration of digital electronics and mechanical design, along with the areas of software development, real-time control, testing and evaluation, and working as part of a team. Teams of students are required to design and construct a fully autonomous computer-controlled mobile robot using LEGO pieces, and compete against other robots at the end of the semester.

This paper describes an autonomous agent created for the Find Life on Mars competition held in Providence, RI. This competition closely parallels research currently being done here at the Colorado School of Mines in Urban Search and Rescue (USAR). Silver Bullet, an autonomous mobile robot, has been constructed under a 2 year NSF Research Experience for Undergraduates program for aiding rescue attempts in USAR situations. The AI approach was to integrate vision and navigation behaviors with affordances to search for signs of life. This paper reports on the results of the Find Life on Mars Competition and the implementation of intelligent search behaviors for a mobile robot.

Successful navigation is one of the most difficult tasks for a mobile robot to undertake. In this paper, we survey a number of published mobile robot navigation schemes that make use of ultrasonic sensors. Ultrasonic sensors were selected as the basis for comparison because of their widespread availability and relatively low cost. The specific approach for this survey is to summarize the published navigation methods that appear to have the most applicability to use in a small mobile robot. This includes summarizing the strengths and weaknesses of the approach, as described by the authors, and assigning relative weights to each method. The highest-ranking methods are selected for simulation to further differentiate between the relative performances of each method. Three of the most promising methods are then implemented on a mobile robot and tested. Finally, we have included our recommendations concerning the suitability of each of the implemented navigation methods for a particular task.

This paper describes a design philosophy developed and currently applied at the University of Alberta ADVANCED ROBOTICS AND TELEOPERATION (ART) Lab. When the philosophy, termed the `Robotic Internet Platform' (RIP) design paradigm, is applied, the result is a teleoperated robot that can be considered an Internet station with the ability to move. Multiple users, with varying roles in system operation, can potentially be connected to the telerobotic system simultaneously. Although the concept is simple, the design implications are great. This paper is intended as a starting-point for developers of telerobotic systems and is meant to aid in several aspects of the design process. It includes considerations that should be made paramount when developing the user interface and the workcell control engine. Two development platforms, TCL/TK and Java, are examined for suitability of use in telerobotic applications.

Smart man machine interfaces turn out to be a key technology for automation applications in industrial environments as well as in scenarios for space applications. For either field the use of Virtual Reality (VR) techniques shows a great potential. At the IRF, a Virtual Reality system was developed and implemented during the last three years which allows the intuitive control of a multi-robot system and different automation systems under one unified VR framework. The general aim of the development was to provide the general framework for Projective Virtual Reality which allows projection of actions that are carried out by users in the virtual world into the real world with the help of robots.

Operating teleoperators in hazardous applications requires expertise of an unseen workspace, and performance must be error-free despite the limited sensory feedback. In emergencies, the most skilled teleoperation personnel may be off-site where there is no specific hook-up for teleoperation. In this work we demonstrate an intuitive gestural interface for teleoperation as opposed to the joystick, console or teach-pendant modes. This can be deployed to control the manipulators from remote sites where only a video communication channel is available. Two modes of control are implemented: (1) Master-Slave mode, where the robot is a model of the human arm, and (2) Pointing mode, where a motion direction is indicated to any type of robotic device. Pointing Gestures are more accurate and can be used for a wider range of devices. This approach enables distributed operation of a wide range of devices and will allow greater flexibility in the operation of sensitive facilities.

The objective of this effort is to evaluate the utility of a teach-by-showing (TBS) technique that learns from the methods and strategies of a human teacher and produces a sensor-driven program that can control a robot to perform the demonstrated task. A teaching gripper was constructed containing both 6 DOF force/torque and position/orientation sensors. The operator picks-up the gripper and demonstrates several successful methods of accomplishing the task using the gripper. The system then generates a procedural program in the language appropriate to the robot host development environment. If during performance of any task, the robot is unable to continue, the operator may intervene by again demonstrating how to perform that portion of the task. This technology will permit personnel who are not skilled in conventional robot programming languages to develop and use sensor-driven robots.

The development of intelligent, autonomous robotic systems requires a twofold approach. On the one hand, capabilities of robots have to be enhanced that they can act and react more autonomously, but on the other hand, the development of a suitable man-machine interface has to be pushed further to be able to control the system. Experiences in controlling a multi-robot-system which is capable of performing different experiment handling and service tasks in a space laboratory environment showed that the realization of a man-machine interface based on advanced virtual reality (VR) techniques is a promising approach for a new command and supervision interface that is intuitively operable. The general aim of the development is to provide the framework for "Projective Virtual Reality" which allows the projection of actions that are carried out by users in the virtual world into the real world with the help of robots. This framework is described in greater detail in a companion paper, whereas this paper concentrates on the exact kinematic and dynamic simulation and animation of human-shaped avatars which act in the virtual environment as representations of users working in the virtual world. The avatars can walk forward, backward and can turn; they can handle tools with their hands, and they can even stoop to pick something up from the floor.

The accepted robot model allows the complete lack of information about the real forms and values of external disturbances and the robot internal dynamics. A qualitative robot property enabling existence of a control which can be generated under such a lack of information and which can simultaneously steer the error vector to the origin in a finite time is natural trackability. A necessary condition for natural trackability is used to generate such a control-natural tracking control. Control algorithms are proved by using the Physical Continuity and Uniqueness Principle established in the paper with or without a combination with the Lyapunov method. The algorithms guarantee different forms of global exponential stability of the robot desired motion and global exponential tracking of any of its desired motions.

Resolved acceleration motion control using the pseudoinverse of the Jacobian matrix is a well-known algorithm for control of redundant robotic manipulators which is sometimes unstable. While the effects of this instability have been discussed in the literature, the cause has not. In this paper the cause of the instability is analyzed, and the unstable growth of joint velocities and accelerations is characterized in terms of the smallest singular value of the Jacobian matrix J of the kinematic function. New stabilizing modifications to the control scheme are derived which are not based on the simple addition of a kinematic component but rather attack the cause of the instability directly. Simulations which validate the new control laws are presented.

In this paper we present the results of an effort to develop an off-line application programming interface (API) for the Robotics Research Corporation (RRC) 7-DOF 1207i robot arm. We have demonstrated the feasibility of a parallel port protocol between two PC's, one running a real-time operating system, performing servo control and trajectory generation, and a second PC running a master program written in a high-level language, performing sensory interaction and task-level control of the robot arm. The protocol developed has proven capable of handling multiple sensory inputs, including voice control, a six-axis force/torque sensor, a gripper with finger force sensors, and a vision system. With this system, we have the ability to quickly develop task-level sensory interactive programs in the C programming language while maintaining and fully utilizing the strengths of the existing RRC controller and its pendant-based leach-programming system. We have successfully applied this system in a howitzer loading demonstration at the US Army's Picatinny Arsenal. The demonstration integrates teach-programming, force-sensing, visual servoing, gripper operation, and voice control.

Target tracking by model independent visual servo control is achieved by augmenting quasi-Newton trust region control with target prediction. Prediction schemes are shown to be effective despite imprecise knowledge of kinematic and camera models. A rank-one Broyden update law provides on-line estimation of the Jacobian, and a control law is provided in the form of a quasi-Newton trust region approach. Robot control is independent of robot and camera configurations, and is able to reach both static and moving targets.

This paper presents a new approach to parallel, motion planning for industrial robot arms with six degrees of freedom in an on-line given 3D environment. The method is based on the A*-search algorithm and needs no essential off-line computations. The algorithm works in an implicitly descrete configuration space. Collisions are detected in the cartesian workspace by hierarchical distance computation based on the given CAD model. By decomposing the 6D configuration space into hypercubes and cyclically mapping them onto multiple processing units, a good load distribution can be achieved. We have implemented the parallel motion planner on a workstation cluster with 9 PCs and tested the planner for several benchmark environments. With optimal discretisation, the new approach usually shows linear, and sometimes even superlinear speedups. In on-line provided environments with static obstacles, the parallel planning times are only a few seconds.

In this paper, a new dexterous teleoperated microhand for operation in small insulation chambers is introduced. The microhand consists of three main components: the cross-shaped mounted magnets, the body together with an electronic circuit and batteries, and the fingers made from two thin ribbons of a copper alloy. The fingers are operated by a shape memory alloy (SMA) actuator. The microhand is remotely levitated and moved within the three-dimensional space with a multi-degree-of-freedom by employing magnetic levitation technology. Moreover, the fingers are remotely activated and controlled by the illumination of several LEDs facing the microhand. A method to analyze the force of the fingers is presented. Passive control of the force of the fingers was achieved by controlling the forward currents of the LEDs to regulate their illumination. The results of several experiments are shown to verify the effectiveness of the proposed teleoperated microhand.

The objective of the ALSARM project phase I is to develop a prototype system to gather environmental measurements inside the Advanced Life Support (ALS) project office Biomass Production Chamber (BPC) at Kennedy Space Center (KSC). The ALSARM project is being performed by NASA KSC in conjunction with the University of Central Florida (UCF). A two semester design course at UCF resulted in the basic ALSARM. KSC engineers developed a final design that would meet the ALS scientific and engineering requirements. The ALSARM is complete and is installed in the BPC. The BPC consists of two separate levels, which are used to grow crops in a closed environment. During the course of study, technicians enter the chamber to measure environmental parameters. Entry of personnel into the BPC disturbs the environment in several ways. ALSARM will be an automated method of taking these measurements and will eliminate the personnel entry, reduce the chamber leak rate, and allow more consistent measurements. The ALSARM will integrate state-of-the-art systems in control, mobility, manipulation, information management, and sensor technologies to perform the BPC environmental measurements. The ALSARM is a three Degree-of-Freedom (DOF) robotic system. The ALSARM can operate manually or automatically, moving to preprogrammed points in the BPC. In phase II of the ALSARM project, (starting Fall 1997), the system will be expanded to include a three DOF End Effector (EE) that will incorporate the existing sensors and be able to take plant samples.

The connections between extraterrestrial exploration and the terrestrial interests of the robotics industry are becoming more apparent as the twentieth century draws to a close. Previously, space applications of robotics technology had occupied an insignificant position with respect to the robotics industry. With few exceptions, space robot projects consumed millions of dollars in R&D expenditure, had limited success, and had little relevance to the industry's development. That situation is now changing as national space agencies adapt to new political and economic realities. This paper explores the changing context of space robotics and the concomitant emergence of a high-growth market opportunity in field robotics.