Nowadays wireless communications is an important aspect of mobile robotics. It is common that mobile robots need to establish wireless links to exchange information with other robots, base stations or sensor nodes. And, as in traditional mobile communications small-scale wireless channel fading also occurs in these scenarios. This phenomenon means that the channel gain will vary significantly over small-distances and in a random manner. This degrades both the communication ability of the robots and as a consequence, their overall performance in executing certain tasks. There is therefore a clear need to compensate for this small-scale fading. We could of course compensate small-scale fading in robotic communications using classical diversity techniques. But these diversity techniques were designed for transceivers that either cannot move, or can move but have no control over their position. In the context of robotic communications we can think of mobile robots as transceivers who know their own position and can also control it. This allows us to create a new form of diversity called mobility diversity whose principle is as follows. If the mobile robot experiences a poor channel gain due to a deep fading then it can alter its location by a small amount in order to find a new point with a higher channel gain (note that a low channel gain requires more transmitter energy to achieve the same SNR at the receiver as a high channel gain). Now, the more points the robot explores then the higher is the probability of obtaining a high channel gain but the consumption of mechanical energy also increases. Thus efficient mobility diversity algorithms (MDAs) must be able to deliver high channel gains while simultaneously using a small amount of mechanical energy. In this thesis, we start by simultaneously considering the theoretical aspects of both wireless communications and robotics that underpin this interdisciplinary problem. We then develop intelligent algorithms (MDAs) to solve the maximise channel gain/ minimise mechanical energy challenge while looking at various modifications that can occur i.e., predetermined and adaptive stopping points; MDAs for robots as wireless relays; MDAs that incorporate energy harvesting; and finally optimisation over a continuous search space. In summary, mobility diversity is a relatively new research area in an early stage of development. In this thesis we have developed a comprehensive theory for MDAs that will form the basis for future applications as outlined in chapter 7.

A study has been performed in the design and fabrication of deployable borehole robots into confined spaces. Three robot systems have been developed to perform a visual survey of a subterranean space where for any reason humans could not enter. A 12mm diameter snake arm was designed with a focus on the cable tensions and the failure modes for the components that make the snake arm. An iterative solver was developed to model the snake arm and algorithmically calculate the snake arms optimal length with consideration of the failure modes. A robot was developed to extend the range capabilities of borehole robots using reconfigurable borehole robots based around established actuation and manufacturing techniques. The expected distance and weight requirements of the robot are calculated alongside the forces the robot is required to generate in order to achieve them. The whegged design incorporated into the tracks is also analysed to measure the capability of the robot over rough terrain. Finally, the experiments to find the actual driving forces of the tracks are performed and used to calculate the actual range of the robot in comparison to the target range. The potential of reconfigurable mobile robots for deployment through boreholes is limited by the requirement for conventional gears, motors, and joints. This chapter explores the use of smart materials and innovative manufacturing techniques to form a novel concept of a self-folding robotic joint for a self-assembling robotic system. The design uses shape memory alloys fabricated in laminate structures with heaters to create folding structures.

Vehicular communications occur when two or more vehicles come into range with one another, to share data over wireless media. Its applications are far-reaching, from toll collection to collision avoidance. Rate Adaptation Algorithms (RAAs) in IEEE 802.11 wireless networks maximize throughput by selecting their t ransmission rates among the multiple available' transmission rates based on the time-varying and location-dependent wireless channel conditions. In this thesis, a detailed study is made' on Adaptive Context-Aware Rate Selection (ACARS) algorithm that is efficient for data transfers, improving energy utilization, and also suitable for road safety applications. The goal of ACARS is to implement a RAA th a t can reliably estimate Signal-t-o-Noise-Ratio (SNR) to the Physical (PHY) layer by the integration transmission power, and Access Point (AP) coordination into its design. One of the major challenges of deploying mobile nodes in wireless networks is the power management. ACARS is able to minimize the total transmission power in the presence of propagation phenomena and mobility of vehicles, by rapid estimation of SNR to the PHY layer. Regarding safety applications in vehicular communications, RAAs need to minimize delay. ACARS minimizes the delay by using optimum data rates which reduces tin' network load in order to meet the application requirements. Simulation results confirm that the airtime (delay) as one of the key factors for safety applications is within the range of 100 ms recommended by the Institute of Electrical and Electronic Engineers (IEEE) standard for vehicle safety applications.

Air-to-surface weapons, and the platforms that deliver them, are becoming increasingly automated and remote from operators. The benefits of remotely operated systems have been widely documented, yet the question of accountability remains an issue - both from a legal standpoint and with regards to public opinion. The current focus of the military and industry is to develop weapons with more autonomy and an increased range. This motivation is part of a wider aim for Network Enabled Capability across all military forces. This thesis focuses on one aspect of Network Enabled Capability, the remote re-tasking of air-to-surface weapons in flight. The aim being to explore the potential capabilities of the human operator that may use such a system. Further, this thesis sets out to investigate the differences between using an automated system for re-tasking air-to-surface missiles in flight as opposed to assigning the task to a human operator. A simulation test-bed facility was established to investigate these research aims. The development of this system first required a complete simulation of two air to surface weapon systems, a generic guided bomb, and an extended range missile. These simulation models were integrated into the test-bed facility to allow real-time targeting and firing of the weapon systems in a 3D simulation environment. Two participant trials were carried out to test firstly, the operator terminal designed for task of re-tasking air-to-surface weapons in flight, and secondly, the operator capacity limits when re-tasking multiple air-to-surface weapons against multiple defended targets. These trials found that the system developed was suitable for the task, with interesting results that prove counter to current expectations of operator performance. Further, operator capacity was found to be at a level that can compete with an ideal automated re-tasking system.

Modular robotics has the potential to transform the perception of robotic systems from machines built for specific tasks to multi-purpose tools capable of performing virtually any task. This thesis presents the design, implementation and study of a new self-reconfigurable modular robotic system for use as a research and education platform. The system features a high-speed genderless connector (HiGen), a hybrid module (HyMod), an extensions framework, and a control architecture. The HiGen connector features inter-module communication and is able to join with other HiGen connectors in a manner that allows either side to disconnect in the event of failure. The rapid actuation of HiGen allows connections to be made and broken at a speed that is, to our knowledge, an order of magnitude faster than existing mechanical genderless approaches that feature single-sided disconnect, benefiting the self-reconfiguration time of modular robots. HyMod is a chain, lattice, and mobile hybrid modular robot, consisting of a spherical joint unit that is capable of moving independently and grouping with other units to form arbitrary cubic lattice structures. HyMod is the first module, to our knowledge, that combines efficient single-module locomotion, enabling self-assembly, with the ability for modules to freely rotate within their lattice positions, aiding the self-reconfigurability of large structures. The extension framework is used to augment the capabilities of HyMod units. Extensions are modules that feature specialized functionality, and interface with HyMod units via passive HiGen connectors, allowing them to be un-powered until required for a task. Control of the system is achieved using a software architecture. Based on message routing, the architecture allows for the concurrent use of both centralized and distributed module control strategies. An analysis of the system is presented, and experiments conducted to demonstrate its capabilities. Future versions of the system created by this thesis could see uses in reconfigurable manufacturing, search and rescue, and space exploration.

Swarm robotics systems have the potential to tackle many interesting problems. Their control software is mostly created by ad-hoc development. This makes it hard to deploy swarm robotics systems in real-world scenarios as it is difficult to analyse, maintain, or extend these systems. Formal methods can contribute to overcome these problems. However, they usually do not guarantee that the implementation matches the specification because the system’s control code is typically generated manually. This thesis studies the application of the supervisory control theory (SCT) framework in swarm robotics systems. SCT is widely applied and well established in the man- ufacturing context. It requires the system and the desired behaviours (specifications) to be defined as formal languages. In this thesis, regular languages are used. Regular languages, in the form of deterministic finite state automata, have already been widely applied for controlling swarm robotics systems, enabling a smooth transition from the ad-hoc development currently in practice. This thesis shows that the control code for swarm robotics systems can be automatically generated from formal specifications. Several case studies are presented that serve as guidance for those who want to learn how to specify swarm behaviours using SCT formally. The thesis provides the tools for the implementation of controllers using formal specifications. Controllers are validated on swarms of up to 600 physical robots through a series of systematic experiments. It is also shown that the same controllers can be automatically ported onto different robotics platforms, as long as they offer the required capabilities. The thesis extends and incorporates techniques to the supervisory control theory framework; specifically, the concepts of global events and the use of probabilistic generators. It can be seen as a step towards making formal methods a standard practice in swarm robotics.

The Bayesian methodology is able to deal with a number of challenges in object tracking, especially with uncertainties in the system dynamics and sensor characteristics. However, model complexities can result in non-analytical expressions which require computationally cumbersome approximate solutions. In this thesis computationally efficient approximate methods for object tracking with complex models are developed. One such complexity is when a large group of objects, referred to as a crowd, is required to be tracked. A crowd generates multiple measurements with uncertain origin. Two solutions are proposed, based on a box particle filtering approach and a convolution particle filtering approach. Contributions include a theoretical derivation for the generalised likelihood function for the box particle filter, and an adaptive convolution particle filter able to resolve the data association problem without the measurement rates. The performance of the two filters is compared over a realistic scenario for a large crowd of pedestrians. Extended objects also generate a variable number of multiple measurements. In contrast with point objects, extended objects are characterised with their size or volume. Multiple object tracking is a notoriously challenging problem due to complexities caused by data association. An efficient box particle filter method for multiple extended object tracking is proposed, and for the first time it is shown how interval based approaches can deal efficiently with data association problems and reduce the computational complexity of the data association. The performance of the method is evaluated on real laser rangefinder data. Advances in digital sensors have resulted in systems being capable of accumulating excessively large volumes of data. Three efficient Bayesian inference methods are developed for object tracking when excessively large numbers of measurements may otherwise cause standard algorithms to be inoperable. The underlying mechanics of these methods are adaptive subsampling and the expectation propagation algorithm.

Feedback Linearisation (FL) is a nonlinear control technique that has gained a lot of attention in the past 30 years. Due to its relatively simple synthesis, the use of FL has been investigated particularly in the aerospace community, because aircraft models are often highly nonlinear and a controller is needed that can guarantee good performance over a wide range of operating conditions. However, mathematical models of real-life physical systems always have a level of uncertainty on them, as they are only ever approximations to the real system. In the current literature, the robustness of FL control has been analysed by extensive simulations, which may miss some worst-case combinations of uncertainties in the model. Alternatively, the robustness of the controlled system has been analysed on simplified linear models, using techniques from classical control, which do not well represent the inherently nonlinear dynamics of the system. This thesis contributes to the literature by using more recent techniques for analysis of nonlinear systems to assess robust stability under FL control. We apply advanced robust and nonlinear analysis techniques without the assumption that the controller has direct access to all the states of the system, by including state estimation or sensors in the closed loop for analysis. We also develop an existing analysis technique in the literature to show that a system under approximate FL control does not violate position limits of the actuator, despite uncertainty in the model, improving the rigour of the analysis. This is applied to a high-fidelity model of an aircraft, designed for use in industry and academia.

This thesis presents a new hybrid active and semi-active control method for vibration suppression in flexible structures. The method uses a combination of a semi-active device and an active control actuator situated elsewhere in the structure to suppress vibrations. The key novelty is to use the hybrid controller to enable the semi-active device to achieve a performance as close to a fully active device as possible. This is accomplished by ensuring that the active actuator can assist the semi-active device in the regions where energy is required. Also, the hybrid active and semi-active controller is designed to minimise the switching of the semi-active controller. The control framework used is the immersion and invariance control technique in combination with a sliding mode control. A two degree-of-freedom system with lightly damped resonances is used as an example system. Both numerical and experimental results are generated for this system and then compared as part of a validation study. The experimental system uses hardware-in-the-loop simulation to simulate the effect of both the degrees-of-freedom. The results show that the concept is viable both numerically and experimentally, and improved vibration suppression results can be obtained for the semi-active device that approaches the performance of an active device. To illustrate the effectiveness of the proposed hybrid controller, it is implemented to keep the contact force constant in the pantograph-catenary system of high-speed trains. A detailed derivation is given after which the simulation results are presented. Then a method to design a reduced order observer using an invariant manifold approach is proposed. The main advantage of this approach is that it enables a systematic design approach, and (unlike most nonlinear observer design methods), it can be generalised over a larger class of nonlinear systems. The method uses specific mapping functions in a way that minimises the error dynamics close to zero. Another important aspect is the robustness property which is due to the manifold attractivity: an important feature when an observer is used in a closed loop control system. The observer design is validated using both numerical simulations and hardware-in-the-loop testing. The proposed observer is then compared with a very well known nonlinear observer based on the off-line solution of the Riccati equation for systems with Lipschitz type nonlinearity. In all cases, the performance of the proposed observer is shown to be excellent.

The thesis considers the development of robust output feedback sliding mode controllers for linear time invariant uncertain systems where output information alone is available to the controller. Two approaches to controller design are discussed. The first uses only the plant dynamics and is called static output feedback sliding mode control. It is shown that a quadratically stable sliding motion may be attained for a bounded uncertain system if and only if the system is minimum phase and a particular subsystem triple satisfies the output feedback design criteria. Sliding mode controllers are sensitive to unmatched uncertainty. Hence a robust design is considered which minimises the effect of uncertainty. The second approach is developed for systems which have design difficulties when only the plant dynamics are considered. Extra dynamics are added and the method is called dynamic output feedback sliding mode control. Closed-loop analysis is carried out and stability of the augmented system is observed. Both controllers guarantee a stable sliding motion despite the presence of bounded uncertainty. Finally, two practical uncertain multivariate industrial examples demonstrate the theoretical developments. The first application is a helicopter model. The open loop dynamics have unstable poles with two stable invariant zeros, variations in model parameters and exhibit high levels of cross coupling. A model following sliding mode controller is used to force the plant outputs to track the outputs of an ideal model. Nonlinear simulation results show the practicality of the method. The second application considers the dynamic output feedback sliding mode control of an aircraft model. The system possesses unstable invariant zeros and requires a dynamic output feedback technique. Simulation results are obtained at different operating points to show the effect of unstable invariant zeros. The examples illustrate the benefits of these robust output feedback based sliding mode control developments.