• Refine Query
  • Source
  • Publication year
  • to
  • Language
  • 109
  • 88
  • 5
  • 3
  • 1
  • Tagged with
  • 1350
  • 702
  • 550
  • 434
  • 404
  • 85
  • 76
  • 64
  • 59
  • 55
  • 45
  • 40
  • 38
  • 36
  • 36
  • About
  • The Global ETD Search service is a free service for researchers to find electronic theses and dissertations. This service is provided by the Networked Digital Library of Theses and Dissertations.
    Our metadata is collected from universities around the world. If you manage a university/consortium/country archive and want to be added, details can be found on the NDLTD website.
131

An incremental clustering and associative learning architecture for intelligent robotics

Keysermann, Matthias Ulrich January 2015 (has links)
The ability to learn from the environment and memorise the acquired knowledge is essential for robots to become autonomous and versatile artificial companions. This thesis proposes a novel learning and memory architecture for robots, which performs associative learning and recall of sensory and actuator patterns. The approach avoids the inclusion of task-specific expert knowledge and can deal with any kind of multi-dimensional real-valued data, apart from being tolerant to noise and supporting incremental learning. The proposed architecture integrates two machine learning methods: a topology learning algorithm that performs incremental clustering, and an associative memory model that learns relationship information based on the co-occurrence of inputs. The evaluations of both the topology learning algorithm and the associative memory model involved the memorisation of high-dimensional visual data as well as the association of symbolic data, presented simultaneously and sequentially. Moreover, the document analyses the results of two experiments in which the entire architecture was evaluated regarding its associative and incremental learning capabilities. One experiment comprised an incremental learning task with visual patterns and text labels, which was performed both in a simulated scenario and with a real robot. In a second experiment a robot learned to recognise visual patterns in the form of road signs and associated them with di erent con gurations of its arm joints. The thesis also discusses several learning-related aspects of the architecture and highlights strengths and weaknesses of the proposed approach. The developed architecture and corresponding ndings contribute to the domains of machine learning and intelligent robotics.
132

Biomimetic manipulator control design for bimanual tasks in the natural environment

Smith, Alex January 2016 (has links)
As robots become more prolific in the human environment, it is important that safe operational procedures are introduced at the same time; typical robot control methods are often very stiff to maintain good positional tracking, but this makes contact (purposeful or accidental) with the robot dangerous. In addition, if robots are to work cooperatively with humans, natural interaction between agents will make tasks easier to perform with less effort and learning time. Stability of the robot is particularly important in this situation, especially as outside forces are likely to affect the manipulator when in a close working environment; for example, a user leaning on the arm, or task-related disturbance at the end-effector. Recent research has discovered the mechanisms of how humans adapt the applied force and impedance during tasks. Studies have been performed to apply this adaptation to robots, with promising results showing an improvement in tracking and effort reduction over other adaptive methods. The basic algorithm is straightforward to implement, and allows the robot to be compliant most of the time and only stiff when required by the task. This allows the robot to work in an environment close to humans, but also suggests that it could create a natural work interaction with a human. In addition, no force sensor is needed, which means the algorithm can be implemented on almost any robot. This work develops a stable control method for bimanual robot tasks, which could also be applied to robot-human interactive tasks. A dynamic model of the Baxter robot is created and verified, which is then used for controller simulations. The biomimetic control algorithm forms the basis of the controller, which is developed into a hybrid control system to improve both task-space and joint-space control when the manipulator is disturbed in the natural environment. Fuzzy systems are implemented to remove the need for repetitive and time consuming parameter tuning, and also allows the controller to actively improve performance during the task. Experimental simulations are performed, and demonstrate how the hybrid task/joint-space controller performs better than either of the component parts under the same conditions. The fuzzy tuning method is then applied to the hybrid controller, which is shown to slightly improve performance as well as automating the gain tuning process. In summary, a novel biomimetic hybrid controller is presented, with a fuzzy mechanism to avoid the gain tuning process, finalised with a demonstration of task-suitability in a bimanual-type situation.
133

Real-time object detection using monocular vision for low-cost automotive sensing systems

Katramados, Ioannis January 2013 (has links)
This work addresses the problem of real-time object detection in automotive environments using monocular vision. The focus is on real-time feature detection, tracking, depth estimation using monocular vision and finally, object detection by fusing visual saliency and depth information. Firstly, a novel feature detection approach is proposed for extracting stable and dense features even in images with very low signal-to-noise ratio. This methodology is based on image gradients, which are redefined to take account of noise as part of their mathematical model. Each gradient is based on a vector connecting a negative to a positive intensity centroid, where both centroids are symmetric about the centre of the area for which the gradient is calculated. Multiple gradient vectors define a feature with its strength being proportional to the underlying gradient vector magnitude. The evaluation of the Dense Gradient Features (DeGraF) shows superior performance over other contemporary detectors in terms of keypoint density, tracking accuracy, illumination invariance, rotation invariance, noise resistance and detection time. The DeGraF features form the basis for two new approaches that perform dense 3D reconstruction from a single vehicle-mounted camera. The first approach tracks DeGraF features in real-time while performing image stabilisation with minimal computational cost. This means that despite camera vibration the algorithm can accurately predict the real-world coordinates of each image pixel in real-time by comparing each motion-vector to the ego-motion vector of the vehicle. The performance of this approach has been compared to different 3D reconstruction methods in order to determine their accuracy, depth-map density, noise-resistance and computational complexity. The second approach proposes the use of local frequency analysis of i ii gradient features for estimating relative depth. This novel method is based on the fact that DeGraF gradients can accurately measure local image variance with subpixel accuracy. It is shown that the local frequency by which the centroid oscillates around the gradient window centre is proportional to the depth of each gradient centroid in the real world. The lower computational complexity of this methodology comes at the expense of depth map accuracy as the camera velocity increases, but it is at least five times faster than the other evaluated approaches. This work also proposes a novel technique for deriving visual saliency maps by using Division of Gaussians (DIVoG). In this context, saliency maps express the difference of each image pixel is to its surrounding pixels across multiple pyramid levels. This approach is shown to be both fast and accurate when evaluated against other state-of-the-art approaches. Subsequently, the saliency information is combined with depth information to identify salient regions close to the host vehicle. The fused map allows faster detection of high-risk areas where obstacles are likely to exist. As a result, existing object detection algorithms, such as the Histogram of Oriented Gradients (HOG) can execute at least five times faster. In conclusion, through a step-wise approach computationally-expensive algorithms have been optimised or replaced by novel methodologies to produce a fast object detection system that is aligned to the requirements of the automotive domain.
134

Intelligent control agent for autonomous UAS

Tantrairatn, Suradet January 2016 (has links)
A self reconfiguring autopilot system is presented, which is based on a rational agent framework that integrates decision making with abstractions of sensing and actions for next generation unmanned aerial vehicles. The objective of the new intelligent control system is to provide advanced capabilities of self-tuning control for a new UAS airframe or adaptation for an old UAS in the presence of failures in adverse flight conditions. High-level system performance is achieved through on-board dynamical monitoring and estimation associated with controller switching and tuning by the agent. The agent can handle an untuned autopilot or retune the autopilot when dynamical changes occur due to aerodynamic and on-board system changes. The system integrates dynamical modelling, hybrid adaptive control, model validation, flight condition diagnosis, control performance evaluation through software agent development. An important feature of the agent is its abstractions from real-time measurements and also its abstractions from model based on-board simulation. The agent, while tuning and supervising the autopilot, also performs real-time evaluations on the effects of its actions.
135

Bayesian non-linear system identification and frequency response analysis with application to soft smart actuators

Jacobs, William January 2016 (has links)
Newly emerging classes of next generation soft-smart actuators are set to have a huge impact on the fields of robotics, orthotics and prosthetics due to their lightweight, high-strain and muscle-like properties. Like muscle, these actuators can be used in multiple roles, e.g. both as actuators and brakes, due their variable compliance. One important class of soft actuator is the dielectric elastomer actuator (DEA). However, DEAs are extremely difficult to control due to their non-linear and time varying dynamics. A crucial step in the advancement of this technology is the development of techniques for systems level modelling and analysis, which is the focus of this thesis. In the first part of the thesis, a set of DEAs are identified and analysed using standard methods from the field of system identification, obtaining non-linear autoregressive with exogenous input (NARX) models. These provide a benchmark against which later methods are evaluated. The key novelty in this part is the development of NARX models of DEAs for use in non-linear frequency-domain analysis. This result provides insight for the first time into how a set of similarly fabricated DEAs vary in different ways. A further aspect of DEA behaviour is their unexplained time varying behaviour. The system identification approach used to identify NARX models of DEAs is in a convenient form such that it can be easily extended to cater for this time varying behaviour. There are however very few available methods for the frequency domain analysis of time varying systems. A novel method for time varying frequency domain analysis of NARX systems is developed in this work and applied to the DEAs. The analysis procedure is used to provide insight on how the dynamic behaviour of DEAs change over time. In the second part of the thesis a novel approach to the joint structure detection and parameter estimation of NARX models is developed using a sparse Bayesian method. The Bayesian framework allows for the estimation of posterior distributions over model parameters, characterising the model uncertainty. Analytic solutions are found that describe model uncertainty in the frequency-domain as confidence bounds on both linear and higher order frequency response functions. The sparse Bayesian identification algorithm is applied to the DEA data sets and is used to give the first non-linear dynamic model of DEAs with uncertainty bounds plus the first description of DEA dynamics in the frequency-domain, again with uncertainty bounds.
136

Design, analysis and trajectory tracking control of underactuated mobile capsule robots

Huda, Md Nazmul January 2016 (has links)
The research on capsule robots (capsubots) has received attraction in recent years because of their compactness, simple structure and their potential use in medical diagnosis (e.g. capsule endoscopy), treatment and surgical assistance. The medical diagnostic capability of a capsule endoscope - which moves with the aid of visceral peristalsis - in the GI (gastro-intestinal) tract can be improved by adding propulsion to it e.g. legged, magnetic or capsubot-type propulsion. Driven by the above needs this thesis presents the design, analysis, trajectory tracking control and implementation of underactuated mobile capsule robots. These capsule robots can be modified and used in in-vivo medical applications. Researches on the capsubottype underactuated system focus on the stabilization of the robot and tracking the actuated configuration. However trajectory tracking control of an unactuated configuration (i.e. the robotmotion)was not considered in the literature though it is the primary requirement of any mobile robot and also crucial for many applications such as in-vivo inspection. Trajectory tracking control for this class of underactuated mechanical systems is still an open issue. This thesis presents a strategy to solve this issue. This thesis presents three robots namely a one-dimensional (1D) capsule robot, a 2D capsule robot and a 2D hybrid capsule robot with incremental capability. Two new acceleration profiles (utroque and contrarium) for the inner mass (IM) - internal moving part of the capsule robot - are proposed, analysed and implemented for the motion generation of the capsule robots. This thesis proposes a two-stage control strategy for the motion control of an underactuated capsule robot. A segment-wise trajectory tracking algorithm is developed for the 1D capsule robot. Theoretical analysis of the algorithm is presented and simulation is performed in the Matlab/Simulink environment based on the theoretical analysis. The algorithm is implemented in the developed capsule robot, the experimentation is performed and the results are critically analyzed. A trajectory tracking control algorithm combining segment-wise and behaviour-based control is proposed for the 2D capsule robot. Detailed theoretical analysis is presented and the simulation is performed to investigate the robustness of the trajectory tracking algorithm to friction uncertainties. A 2D capsule robot prototype is developed and the experimentation is performed. A novel 2D hybrid robot with four modes of operation - legless motion mode, legged motion mode, hybrid motion mode and anchoring mode - is also designed which uses one set of actuators in all operating modes. The theoretical analysis, modelling and simulation is performed. This thesis demonstrates effective ways of propulsion for in-vivo applications. The outer-shape of the 1D and 2D capsule robots can be customized according to the requirement of the applications, as the propulsion mechanisms are completely internal. These robots are also hermetically sealable (enclosed) which is a safety feature for the in-vivo robots. This thesis addresses the trajectory tracking control of the capsubot-type robot for the first time. During the experimentation the 1D robot prototype tracks the desired position trajectory with some error (relative mean absolute error: 16%). The trajectory tracking performance for the 2D capsubot improves as the segment time decreases whereas tracking performance declines as the friction uncertainty increases. The theoretical analysis, simulation and experimental results validate the proposed acceleration profiles and trajectory tracking control algorithms. The designed hybrid robot combines the best aspects of the legless and legged motions. The hybrid robot is capable of stopping in a suspected region and remain stationary for a prolonged observation for the in-vivo applications while withstanding the visceral peristalsis.
137

Physics-based character locomotion control with large simulation time steps

Greer, David Andrew January 2016 (has links)
Physical simulated locomotion allows rich and varied interactions with environments and other characters. However, control is di cult due to factors such as a typical character's numerous degrees of freedom and small stability region, discontinuous ground contacts, and indirect control over the centre of mass. Previous academic work has made signi cant progress in addressing these problems, but typically uses simulation time steps much smaller than those suitable for games. This project deals with developing control strategies using larger time steps. After describing some introductory work showing the di culties of implementing a handcrafted controller with large physics time steps, three major areas of work are discussed. The rst area uses trajectory optimization to minimally alter reference motions to ensure physical validity, in order to improve simulated tracking. The approach builds on previous work which allows ground contacts to be modi ed as part of the optimization process, extending it to 3D problems. Incorporating contacts introduces di cult complementarity constraints, and an exact penalty method is shown here to improve solver robustness and performance compared to previous relaxation methods. Trajectory optimization is also used to modify reference motions to alter characteristics such as timing, stride length and heading direction, whilst maintaining physical validity, and to generate short transitions between existing motions. The second area uses a sampling-based approach, previously demonstrated with small time steps, to formulate open loop control policies which reproduce reference motions. As a prerequisite, the reproducibility of simulation output from a common game physics engine, PhysX, is examined and conditions leading to highly reproducible behaviour are determined. For large time steps, sampling is shown to be susceptible to physical inva- lidities in the reference motion but, using physically optimized motions, is successfully applied at 60 time steps per second. Finally, adaptations to an existing method using evolutionary algorithms to learn feedback policies are described. With large time steps, it is found to be necessary to use a dense feedback formulation and to introduce phase-dependence in order to obtain a successful controller, which is able to recover from impulses of several hundred Newtons applied for 0.1s. Additionally, it is shown that a recent machine learning approach based on support vector machines can identify whether disturbed character states will lead to failure, with high accuracy (99%) and with prediction times in the order of microseconds. Together, the trajectory optimization, open loop control, and feedback developments allow successful control for a walking motion at 60 time steps per second, with control and simulation time of 0.62ms per time step. This means that it could plausibly be used within the demanding performance constraints of games. Furthermore, the availability of rapid failure prediction for the controller will allow more high level control strategies to be explored in future.
138

Performance analysis of cross-directional control systems

Taylor, Andrew January 2003 (has links)
No description available.
139

Optimum control of linear systems with norm constraints

Vinter, Richard Bertrand January 1972 (has links)
No description available.
140

Artificial hormone network for adaptable robots

Teerakittikul, Pitiwut January 2013 (has links)
With current robotic technologies, it generally remains unreliable to use fully autonomous robots in high-risk robotic applications such as search and rescue, surveillance or exploration in disaster scenarios. One of the main issues comes from the fact that unstructured real-world environments are dynamic and full of interventions. Therefore, for autonomous robots to operate in such environments, the ability to adapt to both internal and external environmental changes is crucial. Being unable to deal with such changes not only could downgrade the performance of the robots but also potentially cause devastating consequences in risky environments. Looking towards nature, it can be observed that biological organisms can cope well with the dynamic unpredictability of real-world environments. One of the key properties which assist biological organisms is the ability to adapt to changing environments by the utilization of hormones in response to environmental cues. This biological feature provides an inspiration for this research which investigates a novel Artificial Hormone Network architecture in providing adaptability for autonomous robots to deal with both internal and external environmental changes in simulations of unstructured real-world environments. The Artificial Hormone Network architecture proposes a new method which allows constructions and interactions of several hormones in order to provide adaptability for autonomous robots in different application scenarios. Two Artificial Hormone Networks (AHN1 and AHN2) are proposed and investigated in this research. Results from experiments correspondingly report better performance in dealing with considered internal and external environmental changes on a robot implemented with the Artificial Hormone Networks than a robot implemented without them. Another important aspect of the Artificial Hormone Network architecture is the ability to be constructed automatically to provide particular adaptability using Cartesian Genetic Programming. Experiment results show that the construction of Artificial Hormone Networks can be evolved and that this evolved system not only performed to a level of adaptability that was acceptable but actually performed better than the “hand-coded” system.

Page generated in 0.0476 seconds