Introducing robots into human environments requires them to handle settings designed specifically for human size and morphology, however, large, conventional humanoid robots with stiff, high powered joint actuators pose a significant danger to humans. By contrast, “anthropomimetic” robots mimic both human morphology and internal structure; skeleton, muscles, compliance and high redundancy. Although far safer, their resultant compliant structure presents a formidable challenge to conventional control. Here we review, and seek to address, characteristic control issues of this class of robot, whilst exploiting their biomimetic nature by drawing upon biological motor control research. We derive a novel learning controller for discovering effective reaching actions created through sustained activation of one or more muscle synergies, an approach which draws upon strong, recent evidence from animal and humans studies, but is almost unexplored to date in musculoskeletal robot literature. Since the best synergies for a given robot will be unknown, we derive a deliberately simple reinforcement learning approach intended to allow their emergence, in particular those patterns which aid linearization of control. We also draw upon optimal control theories to encourage the emergence of smoother movement by incorporating signal dependent noise and trial repetition. In addition, we argue the utility of developing a detailed dynamic model of a complete robot and present a stable, physics-based model, of the anthropomimetic ECCERobot, running in real time with 55 muscles and 88 degrees of freedom. Using the model, we find that effective reaching actions can be learned which employ only two sequential motor co-activation patterns, each controlled by just a single common driving signal. Factor analysis shows the emergent muscle co-activations can be reconstructed to significant accuracy using weighted combinations of only 13 common fragments, labelled “candidate synergies”. Using these synergies as drivable units the same controller learns the same task both faster and better, however, other reaching tasks perform less well, proportional to dissimilarity; we therefore propose that modifications enabling emergence of a more generic set of synergies are required. Finally, we propose a continuous controller for the robot, based on model predictive control, incorporating our model as a predictive component for state estimation, delay-compensation and planning, including merging of the robot and sensed environment into a single model. We test the delay compensation mechanism by controlling a second copy of the model acting as a proxy for the real robot, finding that performance is significantly improved if a precise degree of compensation is applied and show how rapidly an un-compensated controller fails as the model accuracy degrades.

The temporal structure of neural oscillations has become a widespread hypothetical \mechanism" to explain how neurodynamics give rise to neural functions. Despite the great number of empirical experiments in neuroscience and mathematical and computa- tional modelling investigating the temporal structure of the oscillations, there are still few systematic studies proposing dynamical explanations of how it operates within closed sensorimotor loops of agents performing minimally cognitive behaviours. In this thesis we explore this problem by developing and analysing theoretical models of evolutionary robotics controlled by oscillatory networks. The results obtained suggest that: i) the in- formational content in an oscillatory network about the sensorimotor dynamics is equally distributed throughout the entire range of phase relations; neither synchronous nor desyn- chronous oscillations carries a privileged status in terms of informational content in relation to an agent's sensorimotor activity; ii) although the phase relations of oscillations with a narrow frequency difference carry a relatively higher causal relevance than the rest of the phase relations to sensorimotor coordinations, overall there is no privileged functional causal contribution to either synchronous or desynchronous oscillations; and iii) oscilla- tory regimes underlying functional behaviours (e.g. phototaxis, categorical perception) are generated and sustained by the agent's sensorimotor loop dynamics, they depend not only on the dynamic structure of a sensory input but also on the coordinated coupling of the agent's motor-sensory dynamics. This thesis also contributes to the Coordination Dynam- ics framework (Kelso, 1995) by analysing the dynamics of the HKB (Haken-Kelso-Bunz) equation within a closed sensorimotor loop and by discussing the theoretical implications of such an analysis. Besides, it contributes to the ongoing philosophical debate about whether actions are either causally relevant or a constituent of cognitive functionalities by bringing this debate to the context of oscillatory neurodynamics and by illustrating the constitutive notion of actions to cognition.

Currently the central challenge facing evolutionary robotics is to determine how best to extend the range and complexity of behaviour supported by evolved neural systems. Implicit in the work described in this thesis is the idea that this might best be achieved through devising neural circuits (tractable to evolutionary exploration) that exhibit complementary functional characteristics. We concentrate on two problem domains; locomotion and sequence learning. For locomotion we compare the use of GasNets and other adaptive networks. For sequence learning we introduce a novel connectionist model inspired by the role of dopamine in the basal ganglia (commonly interpreted as a form of reinforcement learning). This connectionist approach relies upon a new neuron model inspired by notions of energy efficient signalling. Two reward adaptive circuit variants were investigated. These were applied respectively to two learning problems; where action sequences are required to take place in a strict order, and secondly, where action sequences are robust to intermediate arbitrary states. We conclude the thesis by proposing a formal model of functional integration, encompassing locomotion and sequence learning, extending ideas proposed by W. Ross Ashby. A general model of the adaptive replicator is presented, incoporating subsystems that are tuned to continuous variation and discrete or conditional events. Comparisons are made with Ross W. Ashby's model of ultrastability and his ideas on adaptive behaviour. This model is intended to support our assertion that, GasNets (and similar networks) and reward adaptive circuits of the type presented here, are intrinsically complementary. In conclusion we present some ideas on how the co-evolution of GasNet and reward adaptive circuits might lead us to significant improvements in the synthesis of agents capable of exhibiting complex adaptive behaviour.

The ability to generate and control assembling structures built over many orders of magnitude is an unsolved challenge of engineering and science. Many of the presumed transformational benefits of nanotechnology and robotics are based directly on this capability. There are still significant theoretical difficulties associated with building such systems, though technology is rapidly ensuring that the tools needed are becoming available in chemical, electronic, and robotic domains. In this thesis a simulated, general-purpose computational prototype is developed which is capable of unlimited assembly and controlled by external input, as well as an additional prototype which, in structures, can emulate any other computing device. These devices are entirely finite-state and distributed in operation. Because of these properties and the unique ability to form unlimited size structures of unlimited computational power, the prototypes represent a novel and useful blueprint on which to base scalable assembly in other domains. A new assembling model of Computational Organization and Regulation over Assembly Levels (CORAL) is also introduced, providing the necessary framework for this investigation. The strict constraints of the CORAL model allow only an assembling unit of a single type, distributed control, and ensure that units cannot be reprogrammed - all reprogramming is done via assembly. Multiple units are instead structured into aggregate computational devices using a procedural or developmental approach. Well-defined comparison of computational power between levels of organization is ensured by the structure of the model. By eliminating ambiguity, the CORAL model provides a pragmatic answer to open questions regarding a framework for hierarchical organization. Finally, a comparison between the designed prototypes and units evolved using evolutionary algorithms is presented as a platform for further research into novel scalable assembly. Evolved units are capable of recursive pairing ability under the control of a signal, a primitive form of unlimited assembly, and do so via symmetry-breaking operations at each step. Heuristic evidence for a required minimal threshold of complexity is provided by the results, and challenges and limitations of the approach are identified for future evolutionary studies.

A current challenge in neuroscience and systems biology is to better understand properties that allow organisms to exhibit and sustain appropriate behaviours despite the effects of perturbations (behavioural robustness). There are still significant theoretical difficulties in this endeavour, mainly due to the context-dependent nature of the problem. Biological robustness, in general, is considered in the literature as a property that emerges from the internal structure of organisms, rather than being a dynamical phenomenon involving agent-internal controls, the organism body, and the environment. Our hypothesis is that the capacity for behavioural robustness is rooted in dynamical processes that are distributed between agent ‘brain', body, and environment, rather than warranted exclusively by organisms' internal mechanisms. Distribution is operationally defined here based on perturbation analyses. Evolutionary Robotics (ER) techniques are used here to construct four computational models to study behavioural robustness from a systemic perspective. Dynamical systems theory provides the conceptual framework for these investigations. The first model evolves situated agents in a goalseeking scenario in the presence of neural noise perturbations. Results suggest that evolution implicitly selects neural systems that are noise-resistant during coupling behaviour by concentrating search in regions of the fitness landscape that retain functionality for goal approaching. The second model evolves situated, dynamically limited agents exhibiting minimalcognitive behaviour (categorization task). Results indicate a small but significant tendency toward better performance under most types of perturbations by agents showing further cognitivebehavioural dependency on their environments. The third model evolves experience-dependent robust behaviour in embodied, one-legged walking agents. Evidence suggests that robustness is rooted in both internal and external dynamics, but robust motion emerges always from the systemin-coupling. The fourth model implements a historically dependent, mobile-object tracking task under sensorimotor perturbations. Results indicate two different modes of distribution, one in which inner controls necessarily depend on a set of specific environmental factors to exhibit behaviour, then these controls will be more vulnerable to perturbations on that set, and another for which these factors are equally sufficient for behaviours. Vulnerability to perturbations depends on the particular distribution. In contrast to most existing approaches to the study of robustness, this thesis argues that behavioural robustness is better understood in the context of agent-environment dynamical couplings, not in terms of internal mechanisms. Such couplings, however, are not always the full determinants of robustness. Challenges and limitations of our approach are also identified for future studies.

Oscillatory neural activity is closely related to cognition and behaviour, with synchronisation mechanisms playing a key role in the integration and functional organization of different cortical areas. Nevertheless, its informational content and relationship with behaviour - and hence cognition - are still to be fully understood. This thesis is concerned with better understanding the role of neuronal oscillations and information dynamics towards the generation of embodied cognitive behaviours and with investigating the efficacy of such systems as practical robot controllers. To this end, we develop a novel model based on the Kuramoto model of coupled phase oscillators and perform three minimally cognitive evolutionary robotics experiments. The analyses focus both on a behavioural level description, investigating the robot's trajectories, and on a mechanism level description, exploring the variables' dynamics and the information transfer properties within and between the agent's body and the environment. The first experiment demonstrates that in an active categorical perception task under normal and inverted vision, networks with a definite, but not too strong, propensity for synchronisation are more able to reconfigure, to organise themselves functionally, and to adapt to different behavioural conditions. The second experiment relates assembly constitution and phase reorganisation dynamics to performance in supervised and unsupervised learning tasks. We demonstrate that assembly dynamics facilitate the evolutionary process, can account for varying degrees of stimuli modulation of the sensorimotor interactions, and can contribute to solving different tasks leaving aside other plasticity mechanisms. The third experiment explores an associative learning task considering a more realistic connectivity pattern between neurons. We demonstrate that networks with travelling waves as a default solution perform poorly compared to networks that are normally synchronised in the absence of stimuli. Overall, this thesis shows that neural synchronisation dynamics, when suitably flexible and reconfigurable, produce an asymmetric flow of information and can generate minimally cognitive embodied behaviours.

In this thesis, novel implementation methods for FPGA based real-time motion control systems are investigated. These methods are examined for conventional and modern controller topologies as well as peripheral device interfaces which are mutually essential pieces of a motion controller. The developed methods are initially tested one by one to assess the performance of the individual design / and finally an assembled solution is developed to test the overall design. Tests of the overall design are realized via hardware-in-the-loop simulation of a real-world control problem, selected as a CNC machining center. The developed methods are discussed in terms of their success, resource consumptions and attainable sampling rates.

This thesis describes the critical game flow and dynamic role assignment based on potential fields in robot soccer game and actions taken depending on role assignment. Role assignment is a standard problem of multi-agent game system like robot soccer and it can be realized by many techniques. In this thesis, game flow is described dynamically in terms of critical zones which is formed by potential fields based on the field environment as hills and valleys.

It is a great advantage of gaining knowledge about the environment for navigation. The environment should be measured during the employment as exactly as possible, in order to be able to react correctly. Robots are being used for different tasks at contaminated environments, places in danger or completely safe places. For robotic use, scanning capabilities of the environment are limited at the present. Adhering to the limitations of the scanning systems, the environment is scanned at a particular level and charted with appropriate software. In some cases, this scanning capability is not enough to react correctly. Taking these needs into consideration, this thesis successfully demonstrates the design and implementation of a scanning platform for mobile robotics. A scanning platform with a laser rangefinder scanner will be addressed in a compact design with maximum field of view for quickly mapping and high resolution. With this design, the scanning system will be provided by rotating laser rangefinder at z-axis in one dimension. Laser rangefinder will also measure the distance as the second dimension of the system. Also it rotates the laser beam at another axis which is different from the measuring distance axis and rotary z-axis. In this way, the scanning system will have the three-dimensional scanning ability. This design allows the scanning platform, continuous scanning capability at three dimensional.

In this thesis, the dynamic performances of kinematically and dynamically adjustable planar mechanisms have been investigated. An adjustable mechanism is here defined to be a mechanism where some of the kinematic and/or dynamic parameters are changed in a controlled manner in order to optimize the dynamic behaviour of the mechanism in spite of variable operating conditions. Here, variable operating conditions refer to variable load(s) on the mechanism and/or variable desired input motion. The dynamic behaviour of the mechanism may be optimized via minimization of the actuator torque/force fluctuations, minimization of energy consumed by the actuators etc. According to the type of the adjustable parameter, the adjustable mechanisms are classified into two groups namely, dynamically adjustable mechanisms and kinematically adjustable mechanisms. Mechanisms, where the main concern is to change a dynamic parameter(s) are called dynamically adjustable mechanisms. In the kinematically adjustable mechanisms, on the other hand , the main concern is to change a kinematic parameter(s). The main objective of this study is to investigate the benefits of adjustable planar mechanisms, regarding different dynamic behaviours under variable operating conditions. To achieve this objective , various simulations have been performed on the computer. In these simulations, practical constraints that will exist in a real application have been taken into account as much as possible. The results reveal that, in many cases, the dynamic behaviour of a planar mechanism may be improved quite extensively via adjustable mechanisms which are obtained from the original mechanisms with slight modifications.