Clarence Wilfred de Silva

Powered prostheses are effective for helping amputees walk on level ground, but these devices are inconvenient to use in complex environments. To address the above issue, the present thesis develops a sensor fusion system in a complete human-prosthesis-environment loop to recognize the environmental context, predict the motion intent of different subjects, and control the motion of the prosthesis.There are challenges for realizing this objective:1. Complex information in the prosthesis-environment loop.2. User-dependent and subjective information in the human-prosthesis loop.3. Noisy decisions.4. Disconnected neural system of the user.5. Infeasibility of continuous control.6. No accommodation of navigation in rough terrains.To resolve the mentioned problems, the present thesis develops several methods. First, the present thesis designs a vision-motion sensor fusion system and a side-view convolutional neural network, to recognize the environmental context. To resolve the user-dependent issue, the present thesis proposes an unsupervised cross-subject adaptation method to recognize the locomotion intent for target subjects at 95% accuracy even if the associated data are not labeled. Then the thesis fuses the sequential decisions based on a hidden Markov model to smooth the decisions. Next, the thesis fuses the sensing and control system of the prosthesis to control the powered prosthesis predictively. To improve the authority of the prosthesis user, the present thesis studies two more challenging areas: 1) Transformation from discrete-state control to continuous-phase control. The present thesis estimates the gait phase of human locomotion in complex terrains with 80% lower error than traditional methods; 2) Transformation from structured terrains to rough terrains. The thesis fuses sequential 3D gaze and the environmental context to predict the foot placements of the wearer in rough terrains. The developed methodology is implemented in a physical prototype, and the performance is validated using both computer simulation and experimentation. The developed system and algorithms can recognize the environmental context and predict the motion intent of different amputee users accurately (at an accuracy = 97%) and efficiently (computing time
View record

Mechatronic devices and multi-domain (multi-physics) systems are widely used in modern industry and other engineering applications. Mechatronic engineering focuses on developing a design solution that integrates multiple domains, particularly electrical and mechanical systems. For a successful product, these systems require to be accurate, fast, reliable, flexible, minimalist, easy to use and cost effective. Such design demands are diverse, can interact with each other, and might be characterized quantitatively, qualitatively, or both. This might require different scales, units, and physical representations between multiple criteria or objectives. Interacting criteria or objectives might be conflicting, e.g., improving one requirement might deteriorate another requirement. This requires reaching a compromise between objectives, a trade-off decision. The present dissertation addresses the multi-objective design optimization problem that involves quantitative and qualitative design criteria and objectives, in a mechatronic system. The methods developed in this thesis are applied to the design of a wearable sleep monitoring system. For the benefit of that application, a design optimization framework is proposed for sensor placement on a human body to improve the wearablity and reliability of a monitoring system that contains the sensors. The developed framework assists the designer in selecting the type and location of the sensors, and the pertinent wiring. The framework uses fuzzy sets and numbers to reduce the subjectivity that arises with qualitative criteria. To describe the qualitative objective comfort, fuzzy measures and the Choquet integral are used, particularly for combining multiple criteria and handling model interactions. Furthermore, fuzzy measures with the Choquet integral and the decision-making method VIKOR are introduced to make a relatively less subjective trade-off decision between conflicting objectives. Finally, a comparison is made between an improved VIKOR method and a fuzzy measure and Choquet integral approach, related to their optimization trade-off decisions. This study leads to a synthesis of all presented results and concludes that the proposed methods provide comparable results and are effective strategies for trade-off decisions. In this manner, the present investigation significantly contributes to the development of a more effective approach for solving a multi-objective design problem with quantitative and qualitative design criteria.

View record

Conceptual design is a crucial phase in the Design Development Process (DDP) of complex mechatronic systems. Yet, the available design support is not adequate for the Conceptual Design Development Process (CDDP), let alone the entire DDP. Typically, linear methodology, called the V-model, is used in the software development life cycle (SDLC) for the DDP. The developed DDP can as well aid in addressing the customer requirements properly according to the degree of detail that is sought. Also, a Conceptual Integrated Model (CIM) can describe products from different viewpoints and can be developed to aid the simulation-based design.The primary focus of the present thesis is the conceptual design phase. The thesis proposes a hierarchical DDP, where the V-model process is expanded into multiple layers. These layers assist in providing increased flexibility to the DDP, in which each design phase is subjected to a separate and independent integration and evaluation. Through this approach, the required functions can be realized, and the lengthy iteration loops, due to incompatible subsystems, are avoided. The second key objective of the present thesis is to develop a CIM for formal concept modeling using the modeling language SysML with generic design functional libraries. The first set of design libraries are the FB libraries, which aid in the development of the functional structure. The second set of design libraries are Amesim simulation software elements, which help establish the concept simulation models. Also, the challenges of the transformation and exchange of information between a descriptive modeling language – SysML – and a multi-physics modeling language – Amesim – are explored. The last key objective of the present thesis is the use of fuzzy measures and fuzzy integrals for the evaluation of the non-functional requirements of the conceptual design phase, where Sugeno lamda-measures are employed to address the uncertainty of the requirements. The present research is conducted starting with a descriptive study for the development of a design concept model, concept simulation, and concept evaluation of an industrial fish cutting machine, which falls into the category of complex mechatronic systems. The evaluation of the approach focuses on improving the quality of the conceptual design.

View record

The present thesis addresses the power management issues in a mobile sensor network, with application in automated water quality monitoring. A water quality monitoring platform typically involves a wireless sensor network (WSN), in which a number of mobile sensor nodes (SN) are deployed in the water body to constantly collect the water-related sensory data such as the dissolved oxygen, pH value, temperature, oxidation-reduction potential, and electrical conductivity. This data is used to compute water quality index values, transmit them via some routing schemes, and eventually make them accessible to the water quality professionals, governing agencies, or the public. Power management is nontrivial in the monitoring of a remote environment, especially when long-term monitoring is anticipated. However, constrained by the limited energy supply and internal characteristics of the devices, without proper power management, the devices may become nonfunctional within the networked monitoring system, and as a consequence, the data or events captured during the monitoring process will become inaccurate or non-transmittable. Research is proposed here to develop three distinct approaches for energy conservation in a sensor network, and apply them for automated monitoring of the quality of water in an extensive and remote aquatic body. This thesis analytically develops and applies several energy efficient schemes for power management in the automated spatiotemporal monitoring of the quality of water in an extensive and remote aquatic environment. In general, the schemes for power management of a sensor network can be investigated from a number of aspects and schemes. Those schemes typically range from physical layer optimization to network layer solutions. Meanwhile, depending on the specific applications, some energy efficient methodologies are custom-designed, and thus have limitations when used in other applications. Given this background, three energy efficient methods are proposed in this thesis for conserving energy within a WSN. Those proposed three methods, including DDASA, Hybrid DPS and GCVD, are studied on both the sensor node level and the system level, which energy-efficiently reduce the energy consumption and save extra energy thereby prolonging the life of the WSN. It is expected that the proposed methods will be applicable in other spatiotemporal monitoring applications.

View record

This dissertation addresses the near-optimal deployment problem of robot-sensory nodes in a spatiotemporal field. With limited resources, monitoring of a complex environment may face serious challenges in providing sufficient information for spatiotemporal signal estimation and reconstruction. It is therefore essential to retrieve most useful information from sampling locations while using a small number of sensor nodes. In this dissertation, three aspects are investigated to overcome the shortcomings of the existing information-based sampling methods.First, a sensor node deployment method is designed to find the minimum number of sensor deployment locations while achieving near-optimal field estimation error. To this end, a sampling-based field exploration method is used to find near-optimal sampling locations over an infinite horizon. Moreover, spatiotemporal correlations of the sampling data are studied to find redundant signals. The corresponding sampling locations of the redundant signals are eliminated concerning the network connectivity.Second, a deep reinforcement learning approach is proposed to accelerate the field exploration. Typically, field exploration methods are heavily dependent on random sampling, which has low efficiency. To avoid unnecessary or redundant sampling locations, observations from the sampling locations are utilized. Then a model-based information gain determination of the sampling locations is developed to evaluate the effectiveness of the approach. The proposed method can determine the informativeness of the spatiotemporal field by learning the information gain from the sampled area. The mobile sensory agents are then encouraged to take more samples in the area of higher information gain. Consequently, the spatiotemporal field can be efficiently explored. Moreover, the selected sampling locations can near-optimally reconstruct the spatiotemporal field using statistical methods.Third, a deep learning framework is designed to provide accurate reconstruction and prediction of the spatiotemporal field, using a limited number of observations. Nonlinear mapping from limited observations to the entire spatiotemporal field is needed in a sufficiently large spatiotemporal field. Hence, a deep learning method is proposed to extract sparse representations of the field and their nonlinear mappings. It is also proven that the proposed framework obeys Lipschitz continuity and that the observations collected by sparse representations are sufficient for spatiotemporal field reconstruction.

View record

Recent advances in the technologies of sensing, robotics, and sensor networks have led to significant progress in environmental telemonitoring. Robotic systems have been widely developed and deployed in the field by using their capabilities of mobile sensing, autonomous navigation, and wireless communication. In particular, robotic monitoring and data sampling at locations of interest may be utilized to characterize and interpret the environmental phenomena of a study area. However, in real-world robotic sensing applications, the limitations of on-board resources will limit the coverage of the monitored area and the extent of acquired data, which will hinder the performance of field estimation and mapping. Meanwhile, the constraints of computational capability of the system components call for a computationally efficient framework to schedule and control the robotic sensing missions.This dissertation investigates and develops systematic sensor scheduling and path planning schemes for environmental field estimation through robotic sampling, and their application in aquatic monitoring. First, a hexagonal grid-based sampling frame is designed to distribute spatially balanced sampling locations over the monitored field. Two novel hexagonal grid-based survey planners are developed to generate energy-efficient sampling paths for the exploratory survey using mobile sensing robots, which can be executed in a computationally efficient manner. Second, an energy-constrained survey planner is developed, which achieves optimal coverage density for sampling, with a limit on the energy budget. The generated survey mission guides the robots to collect data samples for estimation and mapping of an unknown field under a Gaussian Process (GP) model. Third, a hierarchical planning framework with a built-in Gaussian Markov Random Field (GMRF) model is developed to provide informative path planning and adaptive sampling for efficient spatiotemporal monitoring. Fourth, the development of a cost-effective, rapidly deployable, and easily maintainable Wireless Mobile Sensor Network (WMSN) for on-line monitoring of surface water is presented. A novel On-Line Water Quality Indexing (OLWQI) scheme is developed and implemented to interpret the large volume of on-line measurements.The experimental results in the present dissertation demonstrate the effectiveness and efficiency of the proposed planning schemes and their application in aquatic environmental monitoring.

View record

At present, robots are applied to specific situations and needs, so special methods are adopted for determining and controlling their movements. This thesis investigates the attention and navigation control of a mobile robot for carrying out dynamically challenging tasks involving humans. Since the robot environment can be complex, dynamic, and unpredictable, the required capacities are defined as those where the robot is required to pay attention to the needed object, plan the best path to the goal location and avoid obstacles. This thesis makes significant contributions in the development of an integrated model of top-down and bottom-up visual attention with self-awareness, and a biology-inspired method of robot planning and obstacle avoidance for a mobile robot. Specifically, an integrated model of top-down and bottom-up visual attention with self-awareness for robot is developed for selecting the highest saliency area in robot view. For mimicking the human attention processes, a robot self-awareness model with a fuzzy decision making system is developed and utilized, which is an important improvement over the existing robot attention models. Inspired by a mammal’s spatial awareness and navigation capabilities, a path planning method based on biological recognition is proposed for navigation tasks of a mobile robot in an unstructured and dynamic environment. An episodic cognitive map that encapsulates the information of scene perception, state neurons and pose perception is built to realize the environment cognition of the robot. An algorithm of the event sequence planning is presented for real-time navigation using the minimum distance between events. The method can choose the optimal planning path based on the tasks. A visual navigation algorithm that has the scale invariant feature transform (SIFT) feature is presented. By using the SIFT feature information, the horizontal coordinate of the matched feature pairs is considered to achieve the purpose of visual navigation. An approach of obstacle avoidance for visual navigation is presented based on a risk function and feasible paths. It can choose an optimal path for obstacle avoidance and then return to path planning through the interaction with the surrounding environment. The developed methodologies are evaluated using both simulation and experimentation.

View record

Mechatronic systems are widely used in modern manufacturing. The key machinery of a manufacturing system should be reliable, flexible, intelligent, less complex, and cost effective, which indeed are distinguishing features of a mechatronic system. To achieve these goals, continuous or on-demand design improvements should be incorporated rapidly and effectively, which will address new design requirements or resolve existing weaknesses of the original design.With the advances in sensor technologies, wireless communication, data storage, and data mining, machine health monitoring (MHM) has achieved significantcapabilities to monitor the performance of an operating machine. The extensive data from the MHM system can be employed in design improvement of the monitored system. In that context, the present dissertation addresses several challenges in applying MHM in design optimization of a mechatronic system.First, this dissertation develops a systematic framework for continuous design evolution of a mechatronic system with MHM. Possible design weaknesses of themonitored system are detected using the information from MHM. The proposed method incorporates an index to identify a possible design weakness by evaluatingthe performance, detecting failures and estimating the health status of the system.Second, improved approaches of intelligent machine fault diagnosis (IMFD) that can be applied to more general machinery and faults, are presented. Thisdissertation develops an IMFD approach based on deep neural networks (DNN). It uses the massive unlabeled MHM data to learn representative features. Usingvery few items of labeled data, this approach can achieve superior diagnosis performance. The dissertation presents another IMFD approach, which uses the convolutional neural networks (CNN) and sensor fusion and has increased diagnosis accuracy and reliability. The end-to-end learning capability of the two approaches enables diagnosis of fault types or machines for which limited prior knowledge is available.Third, a hierarchical DNN-based method of remaining useful life (RUL) prediction is developed. It achieves high accuracy of RUL prediction by modeling thesystem degradation on different health stages. This method generates a better estimate of the system RUL, which provides accurate information for the evaluation of system design.

View record

This dissertation modifies several estimation distribution algorithms (EDAs) and implements them in engineering optimization problems. The EDAs are population-based evolutionary algorithms, which employ extreme elitism selection. The main work of the present study is outlined below. First, an approach of extreme elitism selection is developed for EDAs. This selection highlights the effect of a few top solutions and advances EDAs to form a primary evolutionary direction. Simultaneously, this selection can also maintain population diversity to make EDAs avoid premature convergence. EDAs with the new selection approach are tested using a set of benchmark low-dimensional and high-dimensional optimization problems. The experimental results show that the EDA based on univariate marginal Gaussian distribution (UMGD) with extreme elitism selection can outperform some other classical evolution algorithms for most problems. Second, the EDA based on UMGD with extreme elitism is implemented for solving the inverse displacement problem (IDP) of a robotic manipulator. This EDA is compared with the EDAs with other selection methods in solving the IDP of a 4-degree-of-freedom (DOF) robotic arm. Next the algorithm is integrated with differential mutation to solve the IDP of a 7-DOF robotic arm. After that, the proposed algorithm is used to search for satisfactory solutions as a continuous curve. The simulation results show this algorithm can reach real time speeds, in practical applications. Third, EDAs based on five different Gaussian distributions are proposed to solve optimization problems with various types of constraints like equality, inequality, linear, nonlinear, continuous or discontinuous. It is found the EDA based on a single multivariate Gaussian distribution with extreme elitism selection can outperform other EDAs. Besides, this EDA has good performance for four engineering design problems. Fourth, EDA is combined with differential mutation to solve multi-objective optimization problems (MOPs). The hybrid algorithm seeks to find the Pareto optimal front for MOPs. EDAs guide the search direction in the evolution while differential mutation keeps a diversified population. A new sampling method that uses more Gaussian models to generate offspring is specially designed for the EDAs for MOPs. In light of no-free-lunch theorem, different probabilistic models and programing codes are adopted for different MOPs.

View record

Image-based localization plays a vital role in many tasks of robotics and computervision, such as global localization, recovery from tracking failure, and loop closuredetection. Recent methods based on regression forests for camera relocalizationdirectly predict 3D world locations for 2D image locations to guide camera poseoptimization. During training, each tree greedily splits the samples to minimizethe spatial variance. This thesis develops techniques to improve the performancecamera pose estimation based on regression forests method and extends its application domains. First, random features and sparse features are combined so thatthe new method only requires an RGB image in the testing. After that, a label-freesample-balanced objective is developed to encourage equal numbers of samplesin the left and right sub-trees, and a novel backtracking scheme is developed toremedy the incorrect 2D-3D correspondence in the leaf nodes caused by greedysplitting. Furthermore, the methods based on regression forests are extended to uselocal features in both training and test stages for outdoor applications, eliminatingtheir dependence on depth images. Finally, a new camera relocalization method isdeveloped using both points and lines. Experimental results on publicly availableindoor and outdoor datasets demonstrate the efficacy of the developed approaches,showing superior or on-par accuracy with several state-of-the-art baselines.Moreover, an integrated software and hardware system is presented for mo-bile robot autonomous navigation in uneven and unstructured indoor environments.This modular and reusable software framework incorporates capabilities of perception and autonomous navigation. The system is evaluated are in both simulationand real-world experiments, demonstrating the efficacy and efficiency of the developed system.

View record

Bilateral teleoperation allows a human operator to interact with a remote environment using the superior actuation and sensing skills of a robot and the unmatched cognitive skills of a human operator. It has shown promising results in applications such as telemedicine, telesurgery, and access to hazardous or remote environments. In all of these applications, the robot has to co-exist with humans and other delicate objects in the environment and therefore has to behave in a compliant (“soft”) manner. Moreover, in order to improve the task performance, the interaction force must be fed back to the operator to feel. In this backdrop, the present thesis focuses on the application of bilateral teleoperation in a homecare environment.In view of the underlying challenges involved with bilateral teleoperation, this dissertation focuses on the development of a complete teleoperation system that can effectively perform in real-time. A primary objective here is to use the impedance control approach to design local controllers for master and slave manipulators where the dynamic relationship between the applied forces and the resulting positions of the manipulators during interaction, is controlled. Impedance control requires the identification of the robot inverse dynamic model that can be computed in real-time and can adapt to changes in the actual dynamics of the robot. A complete data-driven learning-based technique called Locally Weighted Projection Regression (LWPR) is therefore used, which does not assume any a-priori knowledge of the inertial parameters of the robot. Performance of the system is improved by using online estimation of impedance of the unknown environment with which the slave manipulator interacts. A method of admittance control is designed. This method overcomes the shortcomings of the standard impedance control, as observed during experimentation. In the end, a method is developed to improve the transparency and position synchronization of the popular approach of wave-variables, which ensures stability under time delay that is induced by the communication channel during the exchange of information between the master and the slave ends. The effectiveness of the present developments is validated in an environment of homecare robotics, through simulation and experimentation, and the results are discussed.

View record

This thesis addresses the problem of interaction control between robot manipulator and the manipulated object in a homecare project. This project aims to use homecare robots at the elderly or disabled people’s home to provide necessary aid and assistance. The robot manipulator is to be operated in autonomous mode or teleoperation mode. The possible first aid or assistance requires direct interaction between the remote side robot manipulator and the human body. To guarantee the compliant interaction between the manipulator and the human body, impedance control was applied. In impedance control, neither the force nor the actual motion of the manipulator is controlled. The dynamic relationship between the interaction force and the resulting motion is controlled so that the interaction force will be monitored and kept at an acceptable range.To shape the mechanical impedance to any desired value as we wish, the remote side interaction force sensing is required. The interaction force could be sensed by a force sensor. Force sensors have a lot of inherent limitations such as narrow bandwidth, sensing noise, and high cost. To avoid a force sensor due to its limitations, sliding mode observers will be applied to estimate the interaction force. The estimated interaction force will be used in the impedance control algorithms. The observer and controller framework will be formulated and the solvability will be discussed thoroughly. In addition, the proposed approach will be compared with some available approaches to show its advantages over others. Bilateral impedance control will be applied in a teleoperation system. The master side impedance controller is to ensure the robust stability of the teleoperation system. The remote slave side impedance controller is used so that the interaction force will be monitored and kept at some acceptable range. Desired impedance parameters selection will be discussed considering the compromise between robust stability and performance. Also, in order to deal with the uncertainties in operator and environment dynamics, a robust performance guaranteed controller synthesis approach will be proposed. Gain-scheduling control could guarantee the stability and the robust performance under those uncertainties.

View record

This dissertation addresses autonomous navigation of robots in a dynamic environment where the existence of moving and/or unknown objects leads to more serious challenges for the robots than those when operating in a traditional stationary environment. Therefore, the use of learning capabilities to facilitate proper robotic operation in a dynamic environment has become an important research area in the past decade. This dissertation proposes several novel learning-based methods to overcome the shortcomings in the existing approaches of autonomous navigation. Three aspects are addressed in the present work. First, a real-time path planning method is designed for autonomous navigation that can generate a path that avoids stationary and moving obstacles. To this end, learning ability is imparted to the robot. The present framework incorporates the statistical planning approach called probabilistic roadmap (PRM), Q-learning together with regime-switching Markov decision process (RSMDP) due to its beneficial characteristics, to form a robust Q-learning. Consequently, the initial path can be improved through robust Q-learning during interaction with a dynamic environment. Second, motion planning under constraints is investigated. Specifically, a closed-form piecewise affine control law, called piecewise affine-extended linear quadratic regulator (PWA-ELQR), for nonlinear-nonquadratic control problems with constraints is proposed. Through linearization and quadratization in the vicinity of the nominal trajectories, nonlinear-nonquadratic control problems can be approximated to linear-quadratic problems where the closed-form results can be derived relatively easily. Third, people detection is integrated into the autonomous navigation task. A classifier trained by a multiple kernel learning-support vector machine (MKL-SVM) is proposed to detect people in sequential images of a video stream. The classifier uses multiple features to describe a person, and learn its parameter values rapidly with the assistance of multiple kernels. In addition to the methodology development, the present research involves computer simulation and physical experimentation. Computer simulation is used to study the feasibility and effectiveness of the developed methodologies of path planning, motion planning and people detection. The experimentation involves autonomous navigation of a homecare robot system. The performance of the developed system is rigorously evaluated through physical experimentation and is improved by refining the developed methodologies.

View record

Built environments are the most prominent and important part of our material culture.Although they are vital for accommodating the exponentially growing and increasinglyurbanized population under the challenging conditions of severe climatic changes anddestabilized global societies, researchers note that the methods of both their design andconstruction need to be significantly improved. The construction industry is the largestsource of waste and remains inefficient, while the architectural profession is beingchallenged by digital technologies, conflicting paradigms, and adverse market realities.What then are the obstacles in improving buildings? Is it the lack of viable ideas? Or, is itthe social reluctance to accept novel ideas? How can architects be the socially relevantforce contributing viable concepts? This thesis builds upon the current theories assertingthe importance of human behavior and intentionality for understanding built environments,thus considering the complexity of both the technical and cultural circumstances. Itestablishes that although architecture is usually considered as a solid and invariable staticform, it never has been a static shell merely delimiting discrete spaces. Through all times,buildings comprised both fixed structures and adjustable devices, which were theinteractive interfaces between the static structures and the transiency of human action. Thisstudy focuses on rigidly foldable kinetic structures as they exemplify the potentialadvantages and challenges of novel architecture; and they are a logical expansion of thetraditional adjustable architectural elements. For decades, theorists expected kineticarchitecture to address the shortcomings of the traditional buildings. However, solvingfolding kinetic geometries is difficult and is hindered by the unintuitive nature of thecurrent digital tools. Furthermore, kinetic environments challenge the traditionalexpectations of occupants. In response, the present thesis investigates the evolving,influenced by digital technologies, paradigms of public spaces, and human reasoningdrivendesign tools, while incorporating such human-centric considerations as socialdynamics, history, and culture into the engineering and architectural design methods forbuilt environments. It is concluded that architecture, its design and construction areprimarily a social endeavor. Therefore understanding the cognitive barriers of design tools and negotiating the social expectations are essential when advancing new technologies forarchitecture.

View record

The bidirectional DC-DC converter is widely used in automobiles, energy storage systems, uninterruptible power supplies and aviation power systems. At present, there are three main problems in this area. The first problem concerns stability of the bidirectional converter when parameters change; the second is maintaining high efficiency of the bidirectional converter over wide load range; the third concerns the sensitivity of the efficiency of the bidirectional converter to parameter changes. This thesis presents a new method to determine the stability of the bidirectional converter using the Lyapunov function method under arbitrary parameter changes. As another new contribution, the stability analysis with eigenvalue method is presented when only the input voltage changes. Although these two methods are used in this thesis to determine the stability of bidirectional dual full bridge DC-DC converter with triple phase-shift control, they can be used to determine the stability of other power converters composed of various power switches and controlled with different control methods. A novel triple phase-shift control method is developed in this thesis to make the bidirectional converter operate at high efficiency and make it robust to parameters changes and output power variations. Simulation results illustrate that the novel control method is better than several other commonly used control methods for the bidirectional converter when component parameters and output power change. The working theory of the bidirectional converter with novel triple phase-shift control method is comprehensively described in the thesis. As another new contribution, the maximum output power of the bidirectional converter is analyzed in detail in the thesis. Simulation studies of this project have provided satisfactory results. Conclusions are made on the presented work and possible future directions in continuing the work are indicated.

View record

This thesis addresses the manipulation control of a mobile robot with the support of a sensor network, for carrying out dynamically challenging tasks. Such tasks are defined as those where the robot is required to first identify objects, approach and grasp the needed objects, and transport them to goal locations in an environment that is dynamic, unstructured and only partially known. In the present work, a robotic system with these capabilities is developed and implemented for use in tasks of search and rescue, and homecare robotics. To this end, this thesis makes significant original contributions in developing a scheme of adaptive nonlinear model predictive control (ANMPC) and a sensor network with dynamic clustering capability for mobile manipulation under challenging conditions. Two object tracking algorithms for color tracking and feature tracking are developed for object identification and tracking. A system that uses Q-learning is developed for mobile robot navigation, which allows the robot to learn and operate in an unknown and unstructured dynamic environment. A traditional approach of image-based visual servo control is developed and demonstrated. The scheme of ANMPC is developed, which incorporates a multi-input multi-output (MIMO) control system that can accommodate constraints, including environmental constraints and physical constraints of the robots. In implementing ANPC scheme, the nonlinear and time-variant model is linearized on line with respect to the current position of the image feature and robot joints, using an adaptive approach. The corresponding control architecture predicts the system outputs and generates optimized control actions according to a cost function. In order to extend the mobile manipulation system to a wider workspace such as that found in cities and home scenarios, a sensor network is designed and developed employing PFSA (Probabilistic Finite State Automata). The developed PFSA is utilized in both modeling the sensor data and organizing and representing the sensor network. An application of object identification and tracking is presented; and a heterogeneous sensor network is developed along with a simulation platform in MATLAB. A self-organized and clustered sensor network, which is based on PFSA, is demonstrated. In conclusion, directions for further research and development are indicated.

View record

Particle Swarm Optimization (PSO) is an evolutionary computation technique, which has been inspired by the group behavior of animals such as schools of fish and flocks of birds. It has shown its effectiveness as an efficient, fast and simple method of optimization. The applicability of PSO in the design optimization of heat sinks is studied in this thesis. The results show that the PSO is an appropriate optimization tool for use in heat sink design.PSO has common problems that other evolutionary methods suffer from. For example, in some cases premature convergence can occur where particles tend to be trapped at local optima and not able to escape in seeking the global optimum. To overcome these problems, some modifications are suggested and evaluated in the present work. These modifications are found to improve the convergence rate and to enhance the robustness of the method. The specific modifications developed for PSO and evaluated in the thesis are: (1) Chaotic Acceleration Factor (2) Chaotic Inertia Factor (3) Global Best Mutation The performance of these modifications is tested through benchmarks problems, which are commonly found and used in the optimization literature. Detailed comparative analysis of the modifications to the classical PSO approach is made, which demonstrates the potential performance improvements. In particular, the modified PSO algorithms are applied to problems with nonlinear constraints. The non-stationary, multi-stage penalty method (PFM) is implemented to handle nonlinear constraints. Pressure vessel optimization and welded beam optimization are two common engineering problems that are used for testing the performance of optimization algorithms and are used here as benchmark testing examples. It is found that the modified PSO algorithms, as developed in this work, outperform many classical and evolutionary optimization algorithms in solving nonlinear constraint problems. The modified PSO algorithm is applied in heat sink design and detailed results are presented. The commercially available software package Ansys Icepak is used in the present work to solve the heat and flow equations in implementing the optimal design variables resulting from the modified PSO algorithms. The main contributions the work are summarized and suggestions are made for possible future work.

View record

This thesis investigates autonomous and fault-tolerant cooperative operation and intelligent control of multi-robot systems in a dynamic, unstructured, and unknown environment. It makes significant original contributions pertaining to autonomous robot cooperation, dynamic task allocation, system robustness, and real-time performance. The thesis develops a fully autonomous and fault tolerant distributed control system framework based on an artificial immune system for cooperative multi-robot systems. The multi-robot system consists of a team of heterogeneous mobile robots which cooperate with each other to achieve a global goal while resolving conflicts and accommodating full and partial failures in the robots. In this framework, the system autonomously chooses the appropriate number of robots required for carrying out the task in an unknown and unpredictable environment. An artificial immune system (AIS) approach is incorporated into the multi-robot system framework, which will provide robust performance, self-deterministic cooperation, and coping with an inhospitable environment. Based on the structure of the human immune system, immune response, immune network theory, and the mechanisms of interaction among antibody molecules, the robots in the team make independent decisions, coordinate, and if required cooperate with each other to accomplish a common goal. As needed for application in cooperative object transportation by mobile robots, the thesis develops a new method of object pose estimation. In this method, a CCD camera, optical encoders, and a laser range finder are the sensors used by the robots to estimate the pose of the detected object. The thesis also develops a market-based algorithm for autonomous multi-robot cooperation, which is then used for comparative evaluation of the performance of the developed AIS-based system framework. In order to validate the developed techniques, a Java-based simulation system and a physical multi-robot experimental system are developed. This practical system is intended to transport multiple objects of interest to a goal location in a dynamic and unknown environment with complex static and dynamic obstacle distributions. The approaches developed in this thesis are implemented in the prototype system in our laboratory and rigorously tested and validated through both computer simulation and physical experimentation.

View record