Pure reinforcement learning does not scale well to domains with many degrees of freedom and particularly to continuous domains. In this thesis, we introduce a hybrid method in which a symbolic planner constructs all approximate solution to a control problem.. Subsequently, a numerical optimisation algorithm is used to refine the qualitative plan into an operational policy. The method is demonstrated on the problem of learning a stable walking gait for a bipedal robot. The contributions of this thesis are as follows. Firstly, the thesis proposes a novel way to generate gait patterns by using a genetic algorithm to generate walking gaits for a humanoid robot using zero moment point as the stability criterion. This is validated on physical robot. Second, we propose an innovative generic learning method that utilises the trainer's domain knowledge about the task to accelerate learning and extend the capabilities of the learning algorithm. The proposed method, which takes advantage of domain knowledge and combines symbolic planning and learning to accelerate and reduce the search space of the learning problem, is tested on a bipedal humanoid robot learning to walk. Finally, it is shown that the extended capability of the learning algorithm handles high complexity learning tasks in the physical world with experimental verification on a physical robot.

Thesis (Ph. D.)--Ohio State University, 2007. / Title from first page of PDF file. Includes bibliographical references (p. 184-194).

Thesis (Ph. D.)--University of Missouri-Columbia, 1996. / Typescript. Vita. Includes bibliographical references (leaves 97-101). Also available on the Internet.

Thesis (Ph. D.)--University of Missouri-Columbia, 1996. / Typescript. Vita. Includes bibliographical references (leaves 97-101). Also available on the Internet.

Thesis (Ph. D.)--Ohio State University, 2007. / Title from first page of PDF file. Includes bibliographical references (p. 139-145).

Presented as a Invited Guest Speaker for Darwin Day

Human bipedal locomotion is characterized by a habitual heel-strike (HS) plantigrade gait, yet the significance of walking foot-posture is not well understood. To date, researchers have not fully investigated the costs of non-heel-strike (NHS) walking. Therefore, we examined walking speed, walk-to-run transition speed, estimated locomotor costs (lower limb muscle volume activated during walking), impact transient (rapid increase in ground force at touchdown) and effective limb length (ELL) in subjects (n=14) who walked at self-selected speeds using HS and NHS gaits. HS walking increases ELL compared with NHS walking since the center of pressure translates anteriorly from heel touchdown to toe-off. NHS gaits led to decreased absolutewalking speeds (P=0.012) and walk-to-run transition speeds (P=0.0025), and increased estimated locomotor energy costs (P<0.0001) compared with HS gaits. These differences lost significance after using the dynamic similarity hypothesis to account for the effects of foot landing posture on ELL. Thus, reduced locomotor costs and increased maximum walking speeds in HS gaits are linked to the increased ELL compared with NHS gaits. However, HS walking significantly increases impact transient values at all speeds (P<0.0001). These trade-offs may be key to understanding the functional benefits of HS walking. Given the current debate over the locomotor mechanics of early hominins and the range of foot landing postures used by nonhuman apes, we suggest the consistent use of HS gaits provides key locomotor advantages to striding bipeds and may have appeared early in hominin evolution.

When considering the array of terrestrial locomotor behaviors, bipedalism is a particularly rare way of moving about the landscape. In fact, humans are the only obligate terrestrial mammalian bipeds. Therefore, understanding both how and why it evolved is particularly intriguing. However, there is debate over why the evolution of bipedalism occurred and there is a large gap in knowledge for the mechanisms that underpin the evolution of these adaptive morphologies. One complicating factor for sorting out which models best explain how our hominin ancestors became bipedal is that they all rely on the same set of traits. Moreover, many of the traits that are thought to be diagnostic of bipedalism are only linked by association and have not been experimentally tested. That is, they do not appear in non-human primates and other quadrupeds. Therefore, addressing why the evolution of bipedalism occurred requires understanding the adaptive significance of traits linked with bipedalism. In this dissertation, I use an experimental approach employing both human and animal models to explore links between morphology and behavior and to tease apart the adaptive significance of particular traits. For the human portion of the dissertation, I use an inverse dynamics approach (estimating muscle forces from kinematic, kinetic, and anatomical data) to determine how modern human anatomy functions while walking using ape-like postures to clarify the links between morphology and energy costs in different mechanical regimes to determine the adaptive significance of postcranial anatomy. The results from this portion of the dissertation suggest that adopting different joint postures results in higher energy costs in humans due to an increase in active muscle volumes at the knee. These results lead to two conclusions important for understanding the evolution of human bipedalism. One is that human anatomy maintains low energy costs of walking in humans compared to chimpanzees regardless of lower limb postures. Second, the results suggest that erect trunk posture may be an important factor in reducing energy costs, therefore indicating that lumbar lordosis (the curvature of the lower spine) is important for reducing costs. For the animal portion of the dissertation, I use rats as a model for the quadrupedal-to-bipedal transition and experimentally induce bipedal posture and locomotion under a variety of loading conditions to determine if traits consistent with the evolution of bipedalism occur and under what conditions. This experimental design also has the ability to determine if there is a role for developmental plasticity in generating bipedal morphology to help answer the question how the evolution of bipedalism occurred. I find that inducing bipedal behaviors in a quadrupedal animal generates morphology consistent with human bipedal traits and that loading conditions have specific effects in different skeletal elements and at particular joints. I also find that there is a plausible role for developmental plasticity in generating adaptive bipedal morphology in the earliest hominins. Overall, the results from the experimental procedures in this dissertation were able to clarify links between behavior and bipedal morphology, demonstrate a plausible role for developmental plasticity in early adaptation to bipedal behavior in australopiths, determine the adaptive significance of human postcranial anatomy, and the ways in which postcranial anatomy reduces costs.

No description available.

Controlling a system with chaotic nature provides the ability to control and maintain orbits of different periods which extends the functionality of the system to be flexible. A system with diverse dynamical behaviours can be achieved. Trajectory flows of chaotic systems can be periodically stabilised using only small perturbations from the controlled parameter. The method of chaos control is the Ott-Grebogi-Yorke method. In non-chaotic systems large system parameters changes are required for performance changes. A sagittal plane biped model which is capable of exhibiting periodic and chaotic locomotion was researched and investigated. The locomotion was either periodic or chaotic depending on the design parameters. Nonlinear dynamic tools such as the Bifurcation Diagram, Lyapunov Exponent and Poincaré Map were used to differentiate parameters which generated periodic motion apart from chaotic ones. Numerical analytical tools such as the Closed Return and Linearization of the Poincaré Map were used to detect unstable periodic orbit in chaotic attractors. Chaos control of the model was achieved in simulations. The system dynamic is of the non-smooth continuous type. Differing from other investigated chaotic systems, the biped model has varying phase space dimensions which can range from 3 to 6 dimensions depending on the phase of walking. The design of the biped was such that its features were anthropomorphic with respect to locomotion. The model, consisting of only the lower body (hip to feet), was capable of walking passively or actively and was manufactured with optimal anthropometric parameters based on ground clearance (to avoid foot scuffing) and basin of attraction simulations. During experimentation, the biped successfully walked down an inclined ramp with minimal aid. Real time data acquisitions were performed to capture the results, and the experimental data of the walking trajectories were analysed and verified against simulations. It was verified that the constructed biped exhibits the same walking trend as the derived theoretical model. / Thesis (M.Sc.Eng.)-University of KwaZulu-Natal, Durban, 2012.