The problem of designing variable structure systems with sliding modes for uncertain continuous-time plants involving mismatched parametric uncertainties and matched uncertainties, nonlinearities and/or disturbances is addressed in this thesis. Sliding mode control and estimation schemes are proposed for this class of plants. Pull and partial state information cases are considered. The latter scenario corresponds to sliding mode controllers using only measurable output signals, and comprises static and dynamic output feedback approaches. The proposed synthesis frameworks are based on linear matrix inequality methods and involve polytopic models. The multi-model paradigm is also explored to study the use of a finite set of Lyapunov matrices instead of a single Lyapunov matrix. Thus, a wider number of systems and control engineering problems can be dealt with. Control strategies using only measurable output signals are proposed for designing a single sliding mode controller for the simultaneous stabilisation of a finite collection of plant models. Design methodologies for sliding mode static and dynamic output feedback controllers based on linear matrix inequalities are described. The problem of state reconstruction using a discontinuous observer with sliding modes for systems with matched and mismatched parametric uncertainties is also studied in this dissertation. The mismatched uncertain component is considered as a disturbance whose effect on the output estimation error has to be minimised. The observer gain is synthesised by solving a convex optimisation problem involving linear matrix inequalities, with a polytopic description of the reduced-order error system, in terms of H performance. A detailed stability analysis is carried out for the sliding mode observer and the class of uncertain systems considered. Throughout this thesis, several design examples illustrate the proposed sliding mode control and estimation schemes, and computer simulations are used to demonstrate their efficacy.

A rudimentary biped activity that is essential in interactive evirtual worlds, such as video-games and training simulations, is stepping. For example, stepping is fundamental in everyday terrestrial activities that include walking and balance recovery. Therefore an effective 3D stepping control algorithm that is computationally fast and easy to implement is extremely valuable and important to character animation research. This thesis focuses on generating real-time controllable stepping motions on-the-fly without key-framed data that are responsive and robust (e.g.,can remain upright and balanced under a variety of conditions, such as pushes and dynami- cally changing terrain). In our approach, we control the character’s direction and speed by means of varying the stepposition and duration. Our lightweight stepping model is used to create coordinated full-body motions, which produce directable steps to guide the character with specific goals (e.g., following a particular path while placing feet at viable locations). We also create protective steps in response to random disturbances (e.g., pushes). Whereby, the system automatically calculates where and when to place the foot to remedy the disruption. In conclusion, the inverted pendulum has a number of limitations that we address and resolve to produce an improved lightweight technique that provides better control and stability using approximate feature enhancements, for instance, ankle-torque and elongated-body.

Robots are widely used as substitutes for humans in situations involving repetitive tasks where a precise and repeatable motion is required. Sports technology is an area which has seen an increase in the implementation of robots which simulate specific human motions required for a sport. One purpose is to test sports equipment, where the requirement is for a motion to be performed with consistent variables. One issue which has arisen frequently in the robot simulation of humans is the inherent presence of vibration excited in a flexible object being manipulated by a robot, and this issue is not unfounded in the situation presented in this research, of a golf robot manipulating a flexible golf club during the simulation of a golf swing. It had been found that during robotic simulations of golf swings performed with the Miyamae Robo V at the Sports Technology Institute at Loughborough University, swing variables such as shaft deformation and clubhead orientation were dissimilar to those measured for human golf swings. Vibrations present in the golf club were identified as the key cause of the disparity between human and robot swing variables. This research sought to address this issue and find a method which could be applied to reduce clubhead vibrations present in robot simulations of a golf swing to improve their similarity to human swings. This would facilitate the use of the golf robot for equipment testing and club fitting. Golf swing variables were studied and measured for 14 human subjects with the aim being to understand the motion that the robot is required to simulate. A vibration damping gripper was then fitted to the robot to test the effect that changing the interface between the robot-excited vibrations and the club would have, this was achieved with a selection of silicone sleeves with differing material properties which could be attached to the club. Preliminary results showed a noticeable reduction in clubhead vibrations and this solution was investigated further. Mathematically modelling the robot was seen as the most suitable method for this as it meant the robot remained functional and allowed a number of solutions to be tested. Several iterations of a mathematical model were developed with the final model being structurally similar to the robot with the addition of a compliant grip and wrist. The method by which the robot is driven was also recognised as having a large effect on the level of vibration excited in the clubhead and the methodology behind generating smooth robot swing profiles is presented. The mathematical model was used to perform 6 swings and the resulting shaft deformation and clubhead vibration were compared with data from human swings. It was found that the model was capable of producing swing variables comparable to human swings, however in the downswing portion of the swing the magnitude of these variables were larger for the simulations. Simulations were made which sought to demonstrate the difference between the model replicating the rigid robot and a compliant system. Reductions in vibration were achieved in all swings, including those driven with robot feedback data which contains oscillations excited by the method with which the robot is driven.

On-off traditional pneumatic drives are most widely used in industry offering low-cost, simple but flexible mechanical operation and relatively high power to weight ratio. For a period of decade from mid 1980's to 1990's, some initiatives were made to develop servo pneumatic drives for most sophisticated applications, employing purpose-designed control valves for pneumatic drives and low friction cylinders. However, it is found that the high cost and complex installation have discouraged the manufacturer from applying servo pneumatic drives to industrial usage, making them less favourable in comparison to their electric counterpart. This research aims to develop low-cost servo pneumatic drives which are capable of point-to-point positioning tasks, suitable for applications requiring intermediate performance characteristics. In achieving this objective, a strategy that involves the use of traditional on-off valve, simple control algorithm and distributed field-bus control networks has been adopted, namely, the design and control of Components-based Integrated Pneumatic Drives (CIPDs).

One of the drawbacks of using current telesurgical devices is the absence of force and tactile feedback from the remote surgical manipulators. The lack of tactile feedback in particular limits the surgeon's perception of tissue characteristics during palpation. This project aims to further the current knowledge about tactile feedback for these applications. Through an extensive literature review the tactile information of importance for the perception of soft tissue features was investigated, concluding that feedback of distributed skin deformation from a compliant finger-like sensor is desired. This is unlike the few existing tactile feedback systems, which use rigid/semi-rigid sensors and detect distributed contact pressure. The specification, suitability and effectiveness of a deformation-based tactile feedback system were subsequently investigated. A detailed set of criteria were compiled for a deformation-based tactile sensor and display, and the design options, issues and existing devices discussed for each. An existing biologically-inspired optical tactile sensor, the TACTIP [Chorley et al., 2009], was modified to encode realistic finger pad deformation information and a remotely-actuated tactile display created to output it on a human operator's finger pad. The final deformation-based system was found to offer a good solution suitable for further testing. Qualitative testing found that the relayed tactile information allowed differentiation between different stiffness tissues and localisation of hard lumps in tissues, even when the encapsulating tissue was stiffer than the sensor. The addition of the system to a teleoperated environment with force and visual feedback greatly improved the lump detection rate from 64% to 98% and the localisation error from 18 mm to 11 mm. Finally, the feedback of bulk shear displacement information in addition to the indentation deformation was explored. Following modification of the tactile system, experiments were conducted to establish its impact on lump localisation. Minimal effect was found, indicating that this information is unnecessary for these interactions. The results of this study are expected to benefit the future development of telesurgical systems for remote palpation.

The work described in this thesis focusses on designing and building two novel physical devices in a robotic arm structure. The arm is intended for human-robot interaction in the domestic assistive robotics area. The first device aims at helping to ensure the safety of the human user. It acts as a mechanical fuse and disconnects the robotic arm link from its motor in case of collision. The device behaves in a rigid manner in normal operational times and in a compliant manner in case of potentially harmful collisions: it relies on a variable compliance. The second device is the end-effector of the robotic arm. It is a novel grasping device that aims at accommodating varying object shapes. This is achieved by the structure of the grasping device that is a soft structure with a compliant and a rigid phase. Its completely soft structure is able to mould to the object's shape in the compliant phase, while the rigid phase allows holding the object in a stable way. In this study, variable compliance is defined as a physical structure's change from a compliant to a rigid behaviour and vice versa. Due to its versatility and effectiveness, variable compliance has become the founding block of the design of the two devices in the robot arm physical structure. The novelty of the employment of variable compliance in this thesis resides in its use in both rigid and soft devices in order to help ensure both safety and adaptable grasping in one integrated physical structure, the robot arm. The safety device has been designed, modelled, produced, tested and physically embedded in the robot arm system. Compared to previous work in this field, the feature described in this thesis' work has a major advantage: its torque threshold can be actively regulated depending on the operational situation. The threshold torque is best described by an exponential curve in the mathematical model while it is best fit by a second order equation in the experimental data. The mismatch is more considerable for high values of threshold torque. However, both curves reflect that threshold torque magnitude increases by increasing the setting of the device. Testing of both the passive decoupling and active threshold torque regulation show that both are successfully obtained. The second novel feature of the robot arm is the soft grasping device inspired by hydrostatic skeletons. Its ability to passively adapts to complex shapes objects, reduces the complexity of the grasping action control. This gripper is low-cost, soft, cable-driven and it features no stiff sections. Its versatility, variable compliance and stable grasp are shown in several experiments. A model of the forward kinematics of the system is derived from observation of its bending behaviour. Variable compliance has shown to be a very relevant principle for the design and implementation of a robotic arm aimed at safely interacting with human users and that can reduce grasp control complexity by passively adapting to the object's shape.

In the past few years there has been a lot of development in the field of tele-presence. These developments have caused tele-presence technologies to become easily accessible and also for the experience to be enhanced. Since tele-presence is not only used for tele-presence assisted group meetings but also in some forms of Computer Supported Cooperative Work (CSCW), these activities have also been facilitated. One of the lingering issues has to do with how to properly transmit presence of non-co-located members to the rest of the group. Using current commercially available tele-presence technology it is possible to exhibit a limited level of social presence but no physical presence. In order to cater for this lack of presence a system is implemented here using tele-operated robots as avatars for remote team members and had its efficacy tested. This testing includes both the level of presence that can be exhibited by robot avatars but also how the efficacy of these robots for this task changes depending on the morphology of the robot. Using different types of robots, a humanoid robot and an industrial robot arm, as tele-presence avatars, it is found that the humanoid robot using an appropriate control system is better at exhibiting a social presence. Further, when compared to a voice only scenario, both robots proved significantly better than with only voice in terms of both cooperative task solving and social presence. These results indicate that using an appropriate control system, a humanoid robot can be better than an industrial robot in these types of tasks and the validity of aiming for a humanoid design behaving in a human-like way in order to emulate social interactions that are closer to human norms. This has implications for the design of autonomous socially interactive robot systems.

In this thesis we describe the design and implementation of a Human Machine Interface (HMI) based on gaze tracking proposed to control robot prostheses. Robot manipulators hold a strong similarity with arm prosthetics, we used a 7 degrees of freedom (DOF) whole arm manipulator to test our HMI in the execution of reaching and grasping tasks. We showed that the interface worked under different control strategies using several velocity profiles. The system was tested by ten subjects with encouraging results. We analysed the performance of the 7-DOF robot manipulator in order to determine the suitability of its application in the development of this project. The original setup of the manipulator worked under joint Proportional and Derivative (PD) control but considering the results of the initial analysis of the system we proposed two alternative control strategies aimed to improve the performance of the manipulator: a feedforward friction compensation technique and joint Proportional Integral and Derivative control (PID). We created a dynamic model of the 7-DOF manipulator in Simmechanics in order to have a better understanding of the system. The friction phenomena of the manipulator was identified, represented through a fitted model and included in the system’s model with the aim of incrementing its accuracy with respect to the real system. The characteristics of the model made it suitable to test and to design control strategies for motion and friction compensation in MATLAB/Simulink. The model of the system was validated using data from the real robot arm and it was used later to tune the PID controllers of the joints of the 7-DOF manipulator using Iterative Feedback Tuning (IFT). Both experimental data and model simulations were used for the tuning procedure considering two different approaches. The data obtained from the friction identification process was used to implement a module for feedforward friction compensation over the pre-configured joint PD control of the manipulator. The responses of the system when using joint PID control and joint PD control with gravity and friction compensation were compared in the execution of motion tasks.

This thesis presents the theoretical development along with verification and validation campaign for fault detection and isolation (FDI) schemes applied to a satellite system. Specifically FDI schemes developed in the thesis are applied to the problems of actuator and sensor fault detection based on the nonlinear model of Mars Express Satellite (MEX). Model based FDI schemes are not implemented in the current health monitor schemes on board the MEX satellite, which is an over actuated system with the actuators and sensors having a specific tetrahedral configuration. A simple nonlinear sliding mode observers based on unit vector and super twisting methods developed in the thesis are verified on both rigid body and flexible mode models of the nonlinear MEX satellite. Faults for thrusters are applied during initial strong controller mode phase and also later stages of sun acquisition mode (SAM). Both dedicated observers and a bank of observers are designed to generate residuals which are used to detect and isolate faults. Unlike the reduced order systems typically obtained when using sliding modes in literature, this particular application produces no reduced order dynamics. All the schemes utilize the equivalent injection signal found in the sliding mode literature, which are then filtered using a simple integrator to generate threshold based residual analysis to detect faults or isolate faulty components. Verification and validation campaign consists initially with specific fault simulation and later with Monte Carlo simulations with uncertainties and noise are used to verify the performance of the designed schemes. A new multi-variable version of the super twisting algorithm is proposed along with a Lyapunov approach to analyze its stability. Existing super twisting schemes assumes a single control input structure. The proposed scheme demonstrates that it can be applicable to vector cases, where the decoupling of the control input between states is not possible. Finally a comparison between first order sliding mode observers based on unit vector methods and a second order super twisting observer is made using Monte Carlo simulations in the presence of noise and uncertainty for both the rigid body and flexible mode models of the satellite. The analysis establishes that both methods show good performance and robustness for fault diagnosis, which can be applied for all controller mode changes during several phases of SAM.

This thesis deals with a robust finite-time adaptive law and its application to estimation and (distributed) control. Through the augmentation of a recently developed novel robust finite-time adaptive law to various schemes such as an adaptive observer (for the case of a single SISO system), an adaptive controller (for the case of a single MIMO system) and a distributed adaptive controller (for the case of multi SIMO and multi MIMO systems), robustness and finite-time convergence in the parameter estimation can be achieved. In contrast to conventional adaptive laws (e.g. gradient descent and least-square based method) which only guarantee exponential stability, robustness is 'achieved without compromising the need for finite-time convergence. Auxiliary filters are constructed exempting the algorithm from requiring state velocity (angular acceleration for the case of robotic manipulator control) in the adaptation algorithm. Capitalizing' on the use of a sliding mode-like switching term in the adaptation law coupled with the introduced auxiliary integrated regressors, the parameters are constructed within finite time. Finite-time convergence for consensus of a distributed cooperative (adaptive) control system is achieved by incorporating finite-time sliding surfaces in each connected agent in a network. Nonnegative matrix theory is extended to allow the Lyapunov analysis of the proposed finite-time distributed adaptive controller for multi-degree multi-manipulators. The algorithms are analysed using Lyapunov analysis to prove stability as well as robustness and finite-time convergence. Lyapunov function for the analysis with the case of bounded disturbance presence is also showcased. The algorithms have been successfully applied to an automotive problem to estimate road gradient and mass of vehicle, requiring engine torque and velocity only. The novel adaptive algorithm has been shown for the practical control of a robotic arm and for the cooperative control of two humanoid robotic manipulators for link and Cartesian coordinate control.