A real-time embedded object classification algorithm is developed through the novel combination of binary feature descriptors, a bag-of-visual-words object model and the cortico-striatal loop (CSL) learning algorithm. The BRIEF, ORB and FREAK binary descriptors are tested and compared to SIFT descriptors with regard to their respective classification accuracies, execution times, and memory requirements when used with CSL on a 12.6 g ARM Cortex embedded processor running at 800 MHz. Additionally, the effect of x2 feature mapping and opponent-color representations used with these descriptors is examined. These tests are performed on four data sets of varying sizes and difficulty, and the BRIEF descriptor is found to yield the best combination of speed and classification accuracy. Its use with CSL achieves accuracies between 67% and 95% of those achieved with SIFT descriptors and allows for the embedded classification of a 128x192 pixel image in 0.15 seconds, 60 times faster than classification with SIFT. X2 mapping is found to provide substantial improvements in classification accuracy for all of the descriptors at little cost, while opponent-color descriptors are offer accuracy improvements only on colorful datasets.
Autonomous vehicles driving on off-road terrain exhibit substantial variation in mobility characteristics even when the
terrain is horizontal and qualitatively homogeneous. This paper presents a simple stochastic model for characterizing
observed variability in vehicle response to terrain and for representing transitions between homogeneous terrain with
local variability or between heterogeneous terrain types. Such a model provides a means for more realistic evaluation of
terrain parameter estimation methods through simulation. A stochastic terrain model in which friction angle and soil
cohesion are represented by Gaussian random variables qualitatively represents observed variability in traction vs. slip
characteristics measured experimentally. The stochastic terrain model is used to evaluate a terrain parameter estimation
method in which terrain forces are first estimated independent of a terrain model, and subsequently, parameters of a
terrain model, such as soil cohesion, friction angle, and stress distribution parameters are determined from estimated
vehicle-terrain forces. Simulation results show drawbar pull vs. slip characteristics resulting from terrain parameter
estimation are within statistical bounds established by the stochastic terrain model.
This paper reports a methodology for inferring terrain parameters from estimated terrain forces in order to allow wheeled
autonomous vehicles to assess mobility in real-time. Terrain force estimation can be used to infer the ability to
accelerate, climb, or tow a load independent of the underlying terrain model. When a terrain model is available, physical
soil properties and stress distribution parameters that relate to mobility are inferred from vehicle-terrain forces using
multiple-model estimation. The approach uses Bayesian statistics to select the most likely terrain parameters from a set
of hypotheses, given estimated terrain forces. The hypotheses are based on the extensive literature of soil properties for
soils with cohesions from 1 - 70 kPa. Terrain parameter estimation is subject to mathematical uniqueness of the net
forces resulting from vehicle-terrain interaction for a given set of terrain parameters; uniqueness properties are
characterized in the paper motivating the approach. Terrain force and parameter estimation requires proprioceptive
sensors - accelerometers, rate gyros, wheel speeds, motor currents, and ground speed. Simulation results demonstrate
efficacy of the method on three terrains - low cohesion sand, sandy loam, and high cohesion clay, with parameter
convergence times as low as .02 sec. The method exhibits an ability to interpolate between hypotheses when no single
hypothesis adequately characterizes the terrain.
This paper develops and demonstrates performance analysis of vibration suppression and damage detection control laws on structures with fatigue cracks. State feedback control laws for the individual tasks of vibration suppression and autonomous damage detection are designed based on low-order models of a damaged structure. These control laws are applied to finite-element models of structures with through-surface and surface cracks. The analysis ascertains the ability of feedback control to enhance sensitivity of modal frequency shifts due to realistic damage and the potential for using the same sensors and actuators for implementing vibration damping control laws that are insensitive to damage. In the control model, damage consists of simple reductions in thickness over a small area of the structure. Finite-element models to which control laws are applied are developed using commercial software (ABAQUS) that more accurately models the crack by releasing element connections or by using line spring elements. Results show that feedback control laws can be designed for either crack detection or vibration suppression using identical hardware. In addition, we demonstrate that simple models of damaged structures are suitable for designing control laws for detecting more complex damage conditions, and we demonstrate the use of commercial software for model-based simulation of controlled structures.
A prevalent method of damage detection is based on identifying changes in modal characteristics due to damage induced variations in stiffness or mass along a structure. It is known that modal frequencies can be insensitive to damage, and the open-loop sensitivity itself depends on modal properties and damage location. Here, we develop methods of designing control laws that enhance the sensitivity of modal characteristics to damage. Sensitivity enhancing control exploits the relationship between control gains and closed-loop dynamics in order to increase the observability of damage. The design methods are based on optimization of cost functions that involve the dependence of classic measures of sensitivity on design variables, which include placement of sensors and actuators and state feedback control gains. Due to the size of the design space and the unknown nature of the cost surface, genetic algorithms are used to find control laws that maximize sensitivity to specific damage types subject to control effort and stability constraints. Optimized control laws designed for sensitivity enhancement of stiffness damage in a cantilevered beam are demonstrated by numerical simulation.
KEYWORDS: Control systems, Feedback control, Actuators, Damage detection, Smart structures, Sensors, Finite element methods, Chemical elements, Signal processing, Optical simulations
In this paper, the use of state feedback control to enhance the sensitivity of modal frequencies to local damage is introduced. The method is intended for smart structures, which embody self-actuation and self-sensing capabilities. A simple example introduces the principle of sensitivity enhancing control for a single degree-of-freedom structure. Then, the method is applied to a finite element model of a cantilevered beam to demonstrate the magnitude of sensitivity enhancement achievable for modest, local damage. Methods of implementing state estimate feedback using point measurements of strain along the beam are described, and initial experiments demonstrating sensitivity enhancement for the cantilevered beam are reported. Results show that enhancement in sensitivity of modal frequencies of vibration to damage can be achieved using a single piezoceramic actuator and multiple piezoelectric strain sensors along the beam. The same sensor-actuator configuration can be used for vibration suppression or other control tasks, providing `dual use' smart structure sensors and actuators.
Sensors are currently available and used to monitor structural performance and loads incurred by bridges already in service. However, there has been limited research concerning the stresses that steel bridge girders endure during transport from the manufacturer to the job site and during the installation process. This paper reports the measured stresses on steel bridge girders during transportation from Lancaster, PA to Hanover, NH and during construction of the Ledyard Bridge on the New Hampshire - Vermont border. Two different monitoring system were developed for this data acquisition in a mobile environment. The first, a fiber optic strain monitoring system, utilizing Bragg grating technology. The second utilized an electrical- resistive foil strain gage network, in conjunction with wireless telemetry equipment. Together, these two systems formed a smart structure system for monitoring bridge girders while confirming the accuracy of data gathered through redundancy. Result conclusively demonstrated for the first time, that stresses in beams during transportation are significant and approach the factor of safety margin in girder design.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.