A human body motion tracking system based on use of the MARG (Magnetic, Angular Rate, and Gravity) sensors has been under development at the Naval Postgraduate School and Miami University. The design of a quaternion-based Kalman filter for processing the MARG sensor data was described in [1]. This paper presents the real-time implementation and testing results of the quaternion-based Kalman filter. Experimental results validate the Kalman filter design, and show the feasibility of the MARG sensors for real-time human body motion tracking.

Real-time tracking of human body motion has applications in tele-operation, synthetic reality and others. A motion tracking system based on use of the MARG sensors has been under development at Naval Postgraduate School and Miami University. The magnetic, angular rate, and gravity (MARG) sensor modules use a combination of three orthogonal magnetometers, three orthogonal angular rate sensors, and three orthogonal accelerometers to measure 3-D orientation of individual limb segments in order to determine posture. This paper presents the latest results of the MARG human body motion tracking system. The design and implementation of a control interface unit (CIU), a real-time 3-D human avatar called "Andy", and a concurrent client-server program are discussed. Experimental testing and evaluation of the overall MARG system is also presented. The system is able to track multiple human limbs in real time. The captured human motion data can be visualized over the Internet by multiple clients using the 3-D avatar.

This presents the latest design and implementation of the magnetic, angular rate, and gravity (MARG) sensor module. The MARG sensor module is designed for measuring 3-DOF orientations in real time without singularities. Each MARG sensor contains orthogonally mounted triads of micromachined rate sensors, accelerometers, and magnetometers for a total of nine sensor components. With an integrated microcontroller, the overall factor is less than one cubic inch. Digital data output rate is 100 Hz. To simplify calibration procedures and filtering algorithms, it is important that the response of the individual sensor components is linear within the typical operating regions. Experiments were conducted utilizing a precision tilt table and results indicate that all the sensor components are linear. A simple hand calibration method that requires no specialized equipment is also described. It was validated by experiments that indicate hand calibration produces results that are nearly equivalent to those obtained following precision tilt table calibration.

This paper presents an improved Kalman filter for real-time tracking of human body motions. An earlier version of the filter was presented at IROS 2001. Since then, the filter has been substantially improved. Real-time tracking of rigid body orientation is accomplished using the MARG (magnetic, angular rate, and gravity) sensors. A MARG sensor measures the three-dimensional local magnetic field, three-dimensional angular rate, and three-dimensional acceleration. A Kalman filter is designed to process measurements provided by the MARG sensors, and to produce real-time orientation represented in quaternions. There are many design decisions as related to choice of state vectors, output equations, process model, etc. The filter design presented in this paper utilizes the Gauss-Newton method for parameter optimization in conjunction with Kalman filtering. The use of the Gauss-Newton method, particularly the reduced-order implementation introduced in the paper, significantly simplifies the Kalman filter design, and reduces computational requirements.

Presents an extended Kalman filter for real-time estimation of rigid body orientation using the newly developed MARG (magnetic, angular rate, and gravity) sensors. Each MARG sensor contains a three-axis magnetometer, a three-axis angular rate sensor, and a three-axis accelerometer. The filter represents rotations using quaternions rather than Euler angles, which eliminates the long-standing problem of singularities associated with attitude estimation. A process model for rigid body angular motions and angular rate measurements is defined. The process model converts angular rates into quaternion rates, which are integrated to obtain quaternions. The Gauss-Newton iteration algorithm is utilized to find the best quaternion that relates the measured accelerations and earth magnetic field in the body coordinate frame to calculated values in the earth coordinate frame. The best quaternion is used as part of the measurements for the Kalman filter. As a result of this approach, the measurement equations of the Kalman filter become linear, and the computational requirements are significantly reduced, making it possible to estimate orientation in real time. Extensive testing of the filter with synthetic data and actual sensor data proved it to be satisfactory. Test cases included the presence of large initial errors as well as high noise levels. In all cases the filter was able to converge and accurately track rotational motions.

Rigid body orientation can be determined without the aid of a generated source using nine-axis MARG (Magnetic field, Angular Rate, and Gravity) sensor unit containing three orthogonally mounted angular rate sensors, three orthogonal linear accelerometers and three orthogonal magnetometers. This paper describes a quaternion-based complementary filter algorithm for processing the output data from such a sensor. The filter forms the basis for a system designed to determine the posture of an articulated body in real-time. In the system the orientation relative to an Earth-fixed reference frame of each limb segment is individually determined through the use of an attached MARG sensor. The orientations are used to set the posture of an articulated body model. Details of the fabrication of a prototype MARG sensor are presented. Calibration algorithms for the sensors and the human body model are also presented. Experimental results demonstrate the effectiveness of the tracking system and verify the correctness of the underlying theory.

Proposes the use of Prolog as a rule-based specification language for coordination of multiple control functions to perform missions with autonomous underwater vehicles. Control of both vehicle motion and mission phase logical sequencing, including specified forms of error recovery, must be accomplished. We define terms used in this type of control system and show that such systems fall into the class of hybrid controllers coupling discrete state/time independent and continuous state/continuous time elements. The software architecture to implement these systems is often composed of three levels for ease of segregation and development of functionality. The article concerns an implementation on the NPS Phoenix vehicle. Embedded in the rule body are functions that interface with the vehicle to gather sensory data and generate signals as required to trigger transitions between control functions, and to initiate commands for control functions. The same code is used for mission specification as for its execution, eliminating the question of correctness. Control of a mission segment using command generation to simultaneously drive the vehicle to a point on space and time is described with experimental results. The development of transitioning signals is discussed alongside error recovery techniques using 'guaranteed phase completion'.

An efficient simulation algorithm for an underwater robotic vehicle with a single manipulator was developed by the authors (1994) which included the hydrodynamic effects due to added mass, viscous drag, fluid acceleration, and buoyancy forces. This work has since been extended to the simulation of more general tree-structured mechanisms having star topologies with a number of different joint types while maintaining the O(N) computational complexity (N is the number of links). Using this new algorithm, this paper describes the development of a real-time dynamic simulation system for underwater robotic systems. The primary goal is the efficient implementation of this general algorithm which has been achieved with C++ through the use of object-oriented design techniques of encapsulation, inheritance, and polymorphism. Coupled with realistic 3D graphical models, a powerful tool results for applications ranging from control system development to on-line displays during deployment. The use of this software system has been demonstrated for a number of systems including Aquarobot, an underwater hexapod under development in Japan for seawall construction and surveying.

In this paper, an efficient dynamic simulation algorithm is developed for an underwater robotic vehicle (URV) with a manipulator. It is based on previous work on efficient O(N) algorithms, where N is the number of links in the manipulator, and has been extended to include the effects of a mobile base (the URV body). In addition, the various hydrodynamic forces exerted on these systems in underwater environments are also incorporated into the simulation. The effects modeled in this work are added mass, viscous drag, fluid acceleration, and buoyancy forces. With efficient implementation of the resulting algorithm, the amount of computation with inclusion of the hydrodynamics is almost double that of the original algorithm for a six degree-of-freedom land-based manipulator with a mobile base. Nevertheless, the amount of computation still only grows linearly with the number of degrees of freedom in the manipulator.>

In this paper, an efficient dynamic simulation algorithm is developed for an unmanned underwater vehicle (UUV) with a robotic manipulator. It is based on an efficient O(N) algorithm where N is the number of links in the manipulator, and has been extended to include the full effects of a mobile base and various hydrodynamic forces that are exerted on these systems in underwater environments. The effects modeled in this paper are added mass, viscous drag, fluid acceleration, and buoyancy forces. With efficient implementation of the resulting algorithm, the amount of computation including hydrodynamics almost doubles over the original algorithm for a six degree-of-freedom land-based manipulator with a mobile base. Nevertheless, the amount of computation still only grows linearly with the number of links in the manipulator.>

Long-range route planning is an important component in the intelligent control system of an autonomous agent. Most attempts to solve it with map data rely on applying simple search strategies to high-resolution, node-and-link representations of the map. These techniques have several disadvantages including large time and space requirements. We present an alternative which utilizes a more intelligent representation of the problem environment. Topographical features are represented as homogeneous-cost regions, greatly reducing storage requirements. Then, the A* search strategy is applied to a dynamically created graph, constructed according to Snell's law. Testing has shown significant speed improvements over competing techniques.

Although the off-road mobility characteristics of wheeled or tracked vehicles are generally recognized as being inferior to those of man and cursorial animals, the complexity of the joint-coordination control problem has thus far frustrated attempts to achieve improved vehicular terrain adaptability through the application of legged locomotion concepts. Nevertheless, the evident superiority of biological systems in this regard has motivated a number of theoretical studies over the past decade which have now reached a state of maturity sufficient to permit the construction of experimental computer-controlled adaptive walking machines. At least two such vehicles are known to have recently demonstrated legged locomotion over smooth hard-surfaced terrain. This paper is concerned with an extension of the present theory of limb coordination for such machines to the case in which the terrain includes regions not suitable for weight-bearing and which must consequently be avoided by the control computer in deciding when and where to successively place the feet of the vehicle. The paper includes a complete problem formalization, a heuristic algorithm for solution of the problem thus posed, and a preliminary evaluation of the proposed algorithm in terms of a computer simulation study.

Although many systems for optical reading of printed matter have been developed and are now in wide use, comparatively little success has been achieved in the automatic interpretation of optical images of three-dimensional scenes. This paper is addressed to the latter problem and is specifically concerned with automatic recognition of aircraft types from optical images. An experimental system is described in which certain features called moment invariants are extracted from binary television images and are then used for automatic classification. This experimental system has exhibited a significantly lower error rate than human observers in a limited laboratory test involving 132 images of six aircraft types. Preliminary indications are that this performance can be extended to a wider class of objects and that identification can be accomplished in one second or less with a small computer.

Critical races can be avoided in the binary encoding of asynchronous sequential machines by making use of general codes employing-multistep changes in secondary variables to accomplish transitions between states. The shortest length code of this type presently available is Huffman's internmeshed row set code, which encodes race-free any 2SO-row flow table with 2So-1 bits. In this paper, a new class of codes called augmented parity check codes (APC codes) is introduced and proposed for race-free encoding. Some of the properties of these codes suggest that minimum codes may require only on the order of So+ log2SObits.

In conventional digital differential analyzers (DDA), the word length used for the transmission of information between integrators is restricted to at most a single magnitude bit and a sign bit. This restriction seriously limits integrator frequency response and has to a large extent been responsible for the failure of DDAs to achieve widespread acceptance as general purpose differential analyzers. In this paper it is shown that DDA speed and accuracy can be greatly improved by using increment word lengths which are approximately one-half the length of integrand registers providing that integration formulas more accurate than Euler integration are used. The programming of such machines for the solution of both linear and nonlinear differential equations is discussed and a quantitative evaluation of performance improvement is presented. At the same time, an effort is made to isolate the principal difficulties in hardware implementation which result from extending the integrator increment resolution.

The detection of hazards in combinational switching networks can be accomplished by Boolean algebra if the independence of component states during transients is taken into account. In this note, several new techniques for carrying out such analysis are introduced. It is also shown that certain ambiguities in previous analysis procedures can lead to incorrect conclusions regarding the presence of dynamic hazards.