Tong, Fung Ling. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2007. / Includes bibliographical references (leaves 94-96). / Abstracts in English and Chinese. / Abstract --- p.i / 摘要 --- p.iii / Acknowledgement --- p.iii / Table of Contents --- p.iv / List of Figures --- p.vii / List of Tables --- p.x / Chapter Chapter 1 - --- Introduction --- p.1 / Chapter Chapter 2 - --- State of the art in Vision-based Localization --- p.6 / Chapter 2.1 --- Extended Kalman Filter-based Localization --- p.6 / Chapter 2.1.1 --- Overview of the EKF algorithm --- p.6 / Chapter 2.1.2 --- Process of the EKF-based localization algorithm --- p.8 / Chapter 2.1.3 --- Recent EKF-based vision-based localization algorithms --- p.10 / Chapter 2.1.4 --- Advantages of the EKF-based localization algorithms --- p.11 / Chapter 2.1.5 --- Disadvantages of the EKF-based localization algorithm --- p.11 / Chapter 2.2 --- Monte Carlo Localization --- p.12 / Chapter 2.2.1 --- Overview of MCL --- p.12 / Chapter 2.2.2 --- Recent MCL-based localization algorithms --- p.14 / Chapter 2.2.3 --- Advantages of the MCL-based algorithm --- p.15 / Chapter 2.2.4 --- Disadvantages of the MCL-based algorithm --- p.16 / Chapter 2.3 --- Summary --- p.16 / Chapter Chapter 3 - --- Vision-based Localization as an Optimization Problem --- p.18 / Chapter 3.1 --- "Relationship between the World, Camera and Robot Body Coordinate System" --- p.18 / Chapter 3.2 --- Formulation of the Vision-based Localization as an Optimization Problem --- p.21 / Chapter 3.3 --- Summary --- p.26 / Chapter Chapter 4 - --- Existing Search Algorithms --- p.27 / Chapter 4.1 --- Overview of the Existing Search Algorithms --- p.27 / Chapter 4.2 --- Search Algorithm for the Proposed Objective Function --- p.28 / Chapter 4.3 --- Summary --- p.30 / Chapter Chapter 5 - --- Proposed Vision-based Localization using Genetic Algorithm --- p.32 / Chapter 5.1 --- Mechanism of Genetic Algorithm --- p.32 / Chapter 5.2 --- Formation of Chromosome --- p.35 / Chapter 5.3 --- Fitness Function --- p.39 / Chapter 5.4 --- Mutation and Crossover --- p.40 / Chapter 5.5 --- Selection and Stopping Criteria --- p.42 / Chapter 5.6 --- Adaptive Search Space --- p.44 / Chapter 5.7 --- Overall Flow of the Proposed Algorithm --- p.46 / Chapter 5.8 --- Summary --- p.47 / Chapter Chapter 6 - --- Experimental Results --- p.48 / Chapter 6.1 --- Test Robot --- p.48 / Chapter 6.2 --- Simulator --- p.49 / Chapter 6.2.1 --- Camera states simulation --- p.49 / Chapter 6.2.2 --- Oscillated walking motion simulation --- p.50 / Chapter 6.2.3 --- Input images simulation --- p.50 / Chapter 6.3 --- Computer for simulations --- p.51 / Chapter 6.4 --- Position and Orientation errors --- p.51 / Chapter 6.5 --- Experiment 1 一 Feature points with quantized noise --- p.53 / Chapter 6.5.1 --- Setup --- p.53 / Chapter 6.5.2 --- Results --- p.56 / Chapter 6.6 --- Experiment 2 一 Feature points added with Gaussian noise --- p.62 / Chapter 6.6.1 --- Setup --- p.62 / Chapter 6.6.2 --- Results --- p.62 / Chapter 6.7 --- Experiment 3 一 Noise reduction performance of the adaptive search space strategy --- p.77 / Chapter 6.7.1 --- Setup --- p.77 / Chapter 6.7.2 --- Results --- p.79 / Chapter 6.8 --- Experiment 4 一 Comparison with benchmark algorithms --- p.83 / Chapter 6.8.1 --- Setup --- p.83 / Chapter 6.8.2 --- Results --- p.85 / Chapter 6.9 --- Discussions --- p.88 / Chapter 6.10 --- Summary --- p.90 / Chapter Chapter 7- --- Conclusion --- p.91 / References --- p.94

Image analysis, and its application to sensory input (computer vision) is a fairly mature field, so it is surprising that its techniques are not extensively used in robotic applications. The reason for this is that, traditionally, robots have been used in controlled environments where sophisticated computer vision was not necessary, for example in car manufacturing. As the field of robotics has moved toward providing general purpose robots that must function in the real world, it has become necessary that the robots be provided with robust sensors capable of understanding the complex world around them. However, when researchers apply techniques previously studied in image analysis literature to the field of robotics, several difficult problems emerge. In this thesis we examine four reasons why it is difficult to apply work in image analysis directly to real-time, general purpose computer vision applications. These are: improvement in the computational complexity of image analysis algorithms, robustness to dynamic and unpredictable visual conditions, independence from domain specific knowledge in object recognition and the development of debugging facilities. This thesis examines each of these areas making several innovative contributions in each area. We argue that, although each area is distinct, improvement must be made in all four areas before vision will be utilised as the primary sensory input for mobile, autonomous robotic applications. In the first area, the computational complexity of image analysis algorithms, we note the dependence of a large number of high-level processing routines on a small number of low-level algorithms. Therefore, improvement to a small set of highly utilised algorithms will yield benefits in a large number of applications. In this thesis we examine the common tasks of image segmentation, edge and straight line detection and vectorisation. In the second area, robustness to dynamic and unpredictable conditions, we examine how vision systems can be made more tolerant to changes of illumination in the visual scene. We examine the classical image segmentation task and present a method for illumination independence that builds on our work from the first area. The third area is the reliance on domain-specific knowledge in object recognition. Many current systems depend on a large amount of hard-coded domainspecific knowledge to understand the world around them. This makes the system hard to modify, even for slight changes in the environment, and very difficult to apply in a different context entirely. We present an XML-based language, the XML Object Definition (XOD) language, as a solution to this problem. The language is largely descriptive instead of imperative so, instead of describing how to locate objects within each image, the developer simply describes the properties of the objects. The final area is the development of support tools. Vision system programming is extremely difficult because large amounts of data are handled at a very fast rate. If the system is running on an embedded device (such as a robot) then locating defects in the code is a time consuming and frustrating task. Many development-support applications are available for specific applications. We present a general purpose development-support tool for embedded, real-time vision systems. The primary case study for this research is that of Robotic soccer, in the international RoboCup Four-Legged league. We utilise all of the research of this thesis to provide the first illumination-independent object recognition system for RoboCup. Furthermore we illustrate the flexibility of our system by applying it to several other tasks and to marked changes in the visual environment for RoboCup itself.

<p>This thesis presents work done within the EC-founded project VISATEC. Due to the different directions of the VISATEC project this thesis has a few different threads.</p><p>A novel presentation scheme for medium level vision features applied to range sensor data and to image sequences. Some estimation procedures for this representation have been implemented and tested. The representation is tensor based and uses higher order tensors in a projective space. The tensor can hold information on several local structures including their relative position and orientation. This information can also be extracted from the tensor.</p><p>A number of well-known techniques are combined in a novel way to be able to perform object pose estimation under changes of the object in position, scale and rotation from a single 2D image. The local feature used is a patch which is resampled in a log-polar pattern. A number of local features are matched to a database and the k nearest neighbors vote an object state parameters. This most probable object states are found through mean-shift clustering.</p><p>A system using multi-cue integration as a means of reaching a higher level of system-level robustness and a higher lever of accuracy is developed and evaluated in an industrial-like-setting. The system is based around a robotic manipulator arm with an attached camera. The system is designed to solve parts of the bin-picking problem. The above mentioned 2D technique for object pose estimation is also evaluated within this system.</p>

<p>Friction Stir Welding (FSW) is a new technology which joins materials by using frictional heat. Inthe first part of this thesis, a profound literature study is performed. The basic principles, therobotic implementation and possibilities to use FSW for high strength titanium alloys areexamined. In the next phase, a FSW-tool is modelled and implemented on an industrial robot in arobot simulation program. Reachability tests are carried out on car body parts and jet engineparts. By using a simulation program with embedded collision detection, all possible weldinglocations are determined on the provided parts. Adaptations like a longer FSW-tool and amodified design are suggested in order to get a better reachability. In different case studies, thenumber of required robots and the reduction of weight and time are investigated and comparedto the current spot welding process.</p>

Previous research in force control has focused on the choice of appropriate servo implementation without corresponding regard to the choice of mechanical hardware. This report analyzes the effect of mechanical properties such as contact compliance, actuator-to-joint compliance, torque ripple, and highly nonlinear dry friction in the transmission mechanisms of a manipulator. A set of requisites for high performance then guides the development of mechanical-design and servo strategies for improved performance. A single-degree-of-freedom transmission testbed was constructed that confirms the predicted effect of Coulomb friction on robustness; design and construction of a cable-driven, four-degree-of- freedom, "whole-arm" manipulator illustrates the recommended design strategies.

Femoroacetabular impingement (FAI) has been increasingly recognized as a cause of early hip osteoarthritis. FAI is characterized by pathologic contact between the femur and acetabular rim during hip join movement, caused by morphological abnormalities. Arthroscopic technique has become increasingly popular for FAI surgical treatment because of its minimal invasiveness. However, it involves cumbersome procedures and over- or under-resection are likely to occur. To tackle this issue, robot-assisted FAI arthroscopy is a well suited approach because it results in high accuracy and reproducible surgical outcomes. This dissertation provides new approaches and methods for the current challenges in the development of robot-assisted FAI arthroscopy. The study has three objectives: 1) to develop a robust calibration method for the A-mode ultrasound probe used for noninvasive bone registration, 2) to develop a bone registration simulator for verifying the registration accuracy and consistency for any given registration point-pattern, and 3) to develop a hip range of motion simulation system that returns the virtual range of motion and determines the bone resection volume. Carefully designed calibration procedures and simulation experiments have been conducted during the study of this research. From the experimental results, the developed ultrasound calibration method successfully reduces the registration errors and is proved to be robust. The results from the registration simulator indicate that the pattern with widely distributed points lead to better registration accuracy and consistency. The hip range of motion simulation system results in acceptable accuracy and successfully generates the resection volume. With further modifications, the ultrasound probe can be successfully calibrated with the developed method, and will be applied for noninvasive bone registration. The registration simulator can also be served as a useful tool for determining the optimized registration point-pattern, which can lead to reduced surgical trauma and registration time. Finally, the developed range of motion simulation system can allow the surgeon to evaluate the surgical outcome and to determine the resection volume even before the surgery begins. To conclude, this dissertation provides useful approaches, methods, and software for developing robot-assisted FAI arthroscopy.

Multi-robot systems have the potential to improve performance through parallelism. Unfortunately, interference often diminishes those returns. Starting from the earliest multi-robot research, a variety of arbitration mechanisms have been proposed to maximize speed-up. Vaughan and his collaborators demonstrated the effectiveness of an arbitration mechanism inspired by biological signalling where the level of aggression displayed by each agent effectively prioritizes the limited resources. But most often these arbitration mechanisms did not do any principled consideration of environmental constraints or task structure, signalling cost and precision of the outcome. These factors have been taken into consideration in this research and a taxonomy of the arbitration mechanisms have been presented. The taxonomy organizes prior techniques and newly introduced novel techniques. The latter include theoretical and practical mechanisms (from minimalist to especially efficient). Practicable mechanisms were evaluated on physical robots for which both data and models are presented. The arbitration mechanisms described span a whole gamut from implicit (in case of robotics, entirely without representation) to deliberately coordinated (via an established Biological model, reformulated from a Bayesian perspective). Another significant result of this thesis is a systematic characterization of system performance across parameters that describe the task structure: patterns of interference are related to a set of strings that can be expressed exactly. This analysis of the domain has the important (and rare) property of completeness, i.e., all possible abstract variations of the task are understood. This research presents efficiency results showing that a characterization for any given instance can be obtained in sub-linear time. It has been shown, by construction, that: (1) Even an ideal arbitration mechanism can perform arbitrarily poorly; (2) Agents may manipulate task-structure for individual and collective good; (3) Task variations affect the influence that initial conditions have on long-term behaviour; (4) The most complex interference dynamics possible for the scenario is a limit cycle behaviour.

The current study evaluated whether robotic guidance can influence movement planning and/or the use of online proprioceptive feedback. Participants were divided into three groups wherein they practiced an aiming task unassisted or via a robotic device that led them through a trajectory with an asymmetric or symmetric velocity profile. Baseline performance was measured prior to training and a post-test included control and tendon vibration trials. Temporal, spatial, and kinematic variables were used to assess planning and online control mechanisms. Results indicated that tendon vibration altered the way individuals planned their movements and used online sensory feedback. Robotic-guided groups appeared to use online feedback to a lesser extent than the unassisted group during tendon vibration trials, based on kinematic data. Individuals may become less inclined to use erroneous proprioceptive feedback following robotic guidance, supporting the potential benefit of robotics in neuro-motor rehabilitation for those with proprioceptive deficits.

There are several advantages to replacing CNC machinery with robotic machine tools and as such robotic machining is emerging into the manufacturing and metal cutting industry. There remain several disadvantages to using robots over CNC stations primarily due to flexibility in robotic manipulators, which severely reduces accuracy when operating under high machining forces. This flexibility is dependent on configuration and thus the configuration can be optimised through posture selection to minimise deflection. In previous work little has been done to account for operating frequency and the additional complications that can arise from frequency dependent responses of robotic machine tools. A Fanuc S-360 manipulator was used to experimentally investigate the benefits of including frequency compensation in posture selection. The robot dynamics first had to be identified and experimental modal analysis was selected due the inherent dependency on frequency characteristics. Specifically, a circle fit operation identified modal parameters and a least squares optimisation generated dynamic parameters for a spatial model. A rigid-link flexible-joint model was selected and a pseudo-joint was used to create an additional DOF to accommodate link flexibility. Posture optimisation was performed using a gradient-descent algorithm that used several starting points to identify a global minimum. The results showed that a subset of modal data that excluded the mode shape vectors could be used to create a model to predict the manipulator vibration response. It was also found that the receptance variation of the manipulator with configuration was insufficient to verify the optimisation throughout the entire selected workspace; however the model was shown to be useful in regions containing multiple peaks where the modelled dynamics were dominant over the highly volatile measured data. Simulations were performed on a redundant planar manipulator to overcome the lack of receptance variation found in the Fanuc manipulator. These simulations showed that there were two mechanisms driving the optimisation; overall amplitude reduction and frequency specific amplitude reduction. Using a stiffness posture measure for comparison, the results of the frequency specific reduction could be separated and were found to be particularly beneficial when operating close to resonant frequencies.

Friction Stir Welding (FSW) is a new technology which joins materials by using frictional heat. Inthe first part of this thesis, a profound literature study is performed. The basic principles, therobotic implementation and possibilities to use FSW for high strength titanium alloys areexamined. In the next phase, a FSW-tool is modelled and implemented on an industrial robot in arobot simulation program. Reachability tests are carried out on car body parts and jet engineparts. By using a simulation program with embedded collision detection, all possible weldinglocations are determined on the provided parts. Adaptations like a longer FSW-tool and amodified design are suggested in order to get a better reachability. In different case studies, thenumber of required robots and the reduction of weight and time are investigated and comparedto the current spot welding process.