High fidelity, compliant robot control requires a sufficiently accurate dynamics model. Often though, it is not possible to obtain a dynamics model sufficiently accurately or at all using analytical methods. In such cases, an alternative is to learn the dynamics model from movement data. This thesis discusses the problems specific to dynamics learning for control under nonstationarity of the dynamics. We refer to the cause of the nonstationarity as the context of the dynamics. Contexts are, typically, not directly observable. For instance, the dynamics of a robot manipulator changes as the robot manipulates different objects and the physical properties of the load – the context of the dynamics – are not directly known by the controller. Other examples of contexts that affect the dynamics are changing force fields or liquids with different viscosity in which a manipulator has to operate. The learned dynamics model needs to be adapted whenever the context and therefore the dynamics changes. Inevitably, performance drops during the period of adaptation. The goal of this work, is to reuse and generalize the experience obtained by learning the dynamics of different contexts in order to adapt to changing contexts fast. We first examine the case that the dynamics may switch between a discrete, finite set of contexts and use multiple models and switching between them to adapt the controller fast. A probabilistic formulation of multiple models is used, where a discrete latent variable is used to represent the unobserved context and index the models. In comparison to previous multiple model approaches, the developed method is able to learn multiple models of nonlinear dynamics, using an appropriately modified EM algorithm. We also deal with the case when there exists a continuum of possible contexts that affect the dynamics and hence, it becomes essential to generalize from a set of experienced contexts to novel contexts. There is very little previous work on this direction and the developed methods are completely novel. We introduce a set of continuous latent variables to represent context and introduce a dynamics model that depends on this set of variables. We first examine learning and inference in such a model when there is strong prior knowledge on the relationship of these continuous latent variables to the modulation of the dynamics, e.g., when the load at the end effector changes. We also develop methods for the case that there is no such knowledge available. Finally, we formulate a dynamics model whose input is augmented with observed variables that convey contextual information indirectly, e.g., the information from tactile sensors at the interface between the load and the arm. This approach also allows generalization to not previously seen contexts and is applicable when the nature of the context is not known. In addition, we show that use of such a model is possible even when special sensory input is not available by using an instance of an autoregressive model. The developed methods are tested on realistic, full physics simulations of robot arm systems including a simplistic 3 degree of freedom (DOF) arm and a simulation of the 7 DOF DLR light weight robot arm. In the experiments, varying contexts are different manipulated objects. Nevertheless, the developed methods (with the exception of the methods that require prior knowledge on the relationship of the context to the modulation of the dynamics) are more generally applicable and could be used to deal with different context variation scenarios.

Our dissertation concerns robotic navigation in dynamic indoor environments using image-based visual homing. Image-based visual homing infers the direction to a goal location S from the navigator’s current location C using the similarity between panoramic images IS and IC captured at those locations. There are several ways to compute this similarity. One of the contributions of our dissertation is to identify a robust image similarity measure – mutual image information – to use in dynamic indoor environments. We crafted novel methods to speed the computation of mutual image information with both parallel and serial processors and demonstrated that these time-savers had little negative effect on homing success. Image-based visual homing requires a homing agent tomove so as to optimise themutual image information signal. As the mutual information signal is corrupted by sensor noise we turned to the stochastic optimisation literature for appropriate optimisation algorithms. We tested a number of these algorithms in both simulated and real dynamic laboratory environments and found that gradient descent (with gradients computed by one-sided differences) works best.

Texte remanié de: Diss. di dottorato--Filosofia del linguaggio--Torino, 1999. / Bibliogr. p. 283-298.

Successful coordination of prehensile action depends upon the selection and control of appropriate reach and grasp movements. This thesis explores how prehensile actions are shaped and regulated by perceptual information. According to ecological psychology, behaviour is achieved through the detection of information specifying the opportunities the environment affords for action. A review of the literature identified that as this information evolves over time, a comprehensive understanding of prehension has to consider how affordance perception and continual guidance of action come together in the pursuit of goal-directed action. In a series of interlinking studies the initiation, hand transport and grasp components of prehension were investigated in order to determine how affordances are manifested in the emergence of, and guidance within, prehensile actions. Study 1 explored the effect of information specifying affordances on the time taken to initiate and perform a ball-posting action. Results indicated that affordance perception was reflected in initiation time, whilst affordance actualisation was reflected in movement time, demonstrating that effects of affordance perception extend beyond action preparation to directly influence the emergence of action. Study 2 investigated the selection and regulation of perceptual information during the guidance of hand transport. General tau theory (Lee, 1998) was applied to test i) whether the ratio of coupling between hand and object motion is held constant throughout the reach, and ii) whether this ratio, k, describes the kinematics of handobject contact under varying spatiotemporal task constraints. Results indicated a constant tau ratio during the middle, but not the end phase of the movement; moreover, although the summary ratio k was not sensitive to task manipulations, the time-dependent counterpart, K(t), did exhibit effects of task constraints. This indicates that the guidance of hand transport was a continuous process, where, dependent upon the task goals, the regulation of perceptual information changed throughout the action. The final study, Study 3, examined digit coordination during the grasp. Focusing on the transition from 2-digit to 3-digit grip configurations, the study addressed whether grip selection is made before or during the action. Results showed the transition between 2- and 3-digit grips occurs at a within trial level. The grip configuration utilised could only be distinguished as 2- or 3-digit during the second half of the movement, indicating that grip selection emerges online during the unfolding action. Together these studies provide evidence of continual guidance of prehensile actions and offer support for the consideration of prehensile action as a set of nested task goals. It was concluded that affordance perception and movement guidance are interrelated and evolve continuously throughout the unfolding prehensile action. In light of these findings, issues relating to action selection emerging online from the organism-environment interaction are discussed.

There is considerable evidence suggesting an innate linkage between the human perceptual and motor systems, which evolve together and assist one another in the production and coordination of movement. A major contributor to this relationship is optic flow, providing movement variables such as navigation, obstacle avoidance, and depth perception. The absence of optic flow leads to the decoupling of perception and action, which has been shown to contribute to decrements in human movement (i.e., negatively impacted locomotion and posture, and slower adaptation to gait perturbations). Despite the importance of maintaining this linkage, optic flow manipulations are often found to be underrepresented in locomotion literature when specifically related to rehabilitation training (i.e., treadmills). This may be a contributor to the lengthy and exhaustive treatment plans. The literature has shown instances where reintroducing optic flow into training protocols has shown larger gait improvements in shorter times than typical ambulation protocols, however, the strength of the perception-action linkage in adulthood is still not well understood and its impact not yet fully explored. Therefore, the current research program aims to fill this gap by evaluating how the reintroduction of optic flow into atypical gait training protocols in both healthy and gait-compromised individuals may provide evidence that could be used to enhance rehabilitative outcomes. This series of conceptually related experiments explores outcome enhancements through neuromuscular level changes (Study One), the recalibration process of perception-action given newly acquired physical constraints (Study Two), and on larger scale gait cycle performances in a rehabilitation setting (Study Three). It is hypothesized that perception- action coupling will lead to increases in neuromuscular elicitation in the absence of voluntary movement (Study One), assist the recalibration process to improve measures of spatial awareness and atypical gait parameters (Study Two), and finally, improve rehabilitative outcomes in a spinal cord injury (SCI) ambulation protocol, both objectively (i.e., gait parameters, dynamic balance, SCI measures) and subjectively (i.e., questionnaires) (Study Three). / Thesis / Master of Science (MSc)

No description available.

No description available.

L'objectif de cette thèse est d'étudier le comportement de traversée d'intersection en conduite automobile sous l'angle des actions visuellement guidées. Dans l'introduction, nous montrons que les principales études portant sur la traversée d'intersection se sont focalisées sur les processus de jugement perceptif ; or nous montrons que la compréhension de cette activité pourrait être affinée en utilisant les concepts de la théorie de la perception directe et ceux issus du contrôle continu. Notre revue de questions nous conduit à rapprocher la tâche de traversée d'intersection à l'intérieur d'un créneau libre du trafic à celle d'interception de cible se déplaçant horizontalement. Or, dans le cadre des tâches d'interception, il a été démontré que la stratégie d'angle de relèvement constant (i.e., Constant Bearing Angle, CBA) a un fort pouvoir explicatif vis-à-vis des données expérimentales. Nous avons donc tiré partie des outils de simulation automobile pour tester la plausibilité de l'utilisation d'une telle stratégie par les conducteurs. Dans le premier chapitre expérimental, nous montrons qu'il est légitime d'étudier le couplage perception-action dans ce type de tâche et que les conducteurs adoptent des régulations comportementales proches de celles observées dans une tâche d'interception de mobile. Le second chapitre expérimental nous permet de mettre en évidence l'utilisation conjointe par les conducteurs d'informations relatives au créneau du trafic (supports globaux) et aux véhicules (supports locaux). / This Ph.D. project aims at studying drivers' road crossing behaviour under the perspective of visually guided actions. In the introduction, we present that former studies related to road crossing behaviour mainly focused on perceptual judgement processes. However, we demonstrate that this kind of task would be better understood with theories from the ecological approach of perception and motor control studies. Our literature review lead us to make a bridge between the task of crossing a road inside a free moving traffic gap and the task of intercepting an horizontally moving target. In this regard, it has been demonstrated that the Constant Bearing Angle strategy (CBA) has a high explanatory power in respect to experimental data. We used driving simulation tools in order to assess the hypothesis of the use of bearing angle based strategy by drivers. In the first experimental chapter, we show that studying perceptual-motor coupling in this task is relevant and that observed behaviour is close to taht observed in horizontally moving target interception tasks. The second experimental chapter demonstrates that drivers' behaviour relies both on visual information related to the moving gap (global contribution) and that related to the independent vehicles (local contributions). In the last experimental chapter, we evidenced that drivers' behaviour does not exclusively rely on the changes of the bearing angle but also on traffic vehicles' optical expansion. We conclude on how our results shed new light on former results obtained with perceptual-judgements based methodology.

La traversée d’intersection constitue une tâche typique dans laquelle plusieurs affordances coexistent, l’une traduisant les possibilités de franchir l’intersection, l’autre traduisant les possibilités d’immobiliser le véhicule afin d’éviter une collision. En nous inspirant des travaux de Gibson et Crooks (1938), nous avons formalisé les affordances Field of Safe Travel (FST) et Minimum Stopping Zone (MSZ) en termes de temps critique pour réussir le franchissement d’intersection et l’immobilisation du véhicule, respectivement. Grâce à un dispositif de réalité virtuelle et un simulateur de conduite, nous avons testé la prise en compte de ces affordances par un agent engagé dans la tâche de référence à travers trois expérimentations. Les résultats, appuyés par notre modèle de prise de décision, montrent que le conducteur intègre non seulement l’affordance FST, mais aussi l’affordance MSZ, dans sa décision de franchir l’intersection. De plus, l’utilisation conjointe de ces deux affordances rend compte des régulations de vitesses produites par un agent devant réaliser la tâche de franchissement en présence d’un véhicule préprogrammée ou soumis au contrôle d’un autre agent. Ces différents résultats montrent que les agents prennent en compte les différentes possibilités qui leur sont offertes (affordances FST et MSZ) pour faire un choix adapté et réguler leur vitesse. / Crossing an intersection is a typical task in which several affordances coexist, one offering the possibilities to safely cross the intersection (the main goal) and another offering the possibilities to stop the driving car to avoid a collision (the alternative goal). Based on Gibson and Crooks’ works (1938), we formalized the so-called Field of Safe Travel (FST) and Minimum Stopping Zone (MSZ) affordances in terms of critical time for safe crossing and for safe stopping, respectively. Using a virtual reality set-up with a driving simulator, three experiments have been designed to test the role of these two affordances on the decision-making and the regulation of action process. Our results, supported by a model of decision-making, reveal that drivers rely not only on the FST affordance, but also on the MSZ affordance, to choose to cross-or-not the intersection. Moreover, the joint use of these affordances is still observed when drivers have to control their velocity to perform the task when either a vehicle with a computed kinematic or a car driven by another agent is approaching. Put together, these results show that drivers take into account several possibilities for action offered by the environment (FST and MSZ affordances) to make a suitable choice and to control their speed when approaching an intersection.

La perception visuelle de l’espace est largement déterminée par les capacités visuelles des individus. Cependant, la recherche sur l’influence de variables dites non visuelles semble indiquer une importance déterminante des dimensions corporelle et émotionnelle sur la perception visuelle. Dans cette thèse, nous nous sommes intéressés à la perception des distances et avons tenté de comprendre en quoi nos perceptions pouvaient refléter la nature de nos interactions sensorimotrices avec notre environnement. Dans une première étude nous avons montré que des contraintes d’actions, tel que le coût énergétique, générées par les dispositions de l’environnement, pouvaient avoir une influence sur la perception de distances allocentriques. Dans une seconde étude, nous avons tenté de mettre en évidence le rôle du processus d’intégration sensorimotrice dans des tâches d’évaluation de distances, ainsi que de comprendre l’influence des modalités non visuelles, telles que l’audition et la motricité, dans ce même type de tâche. Enfin, dans une troisième étude, nous nous sommes intéressés au concept de fluence – i.e., information métacognitive renseignant le système sur la qualité de ses interactions dans son environnement – et proposons l’idée selon laquelle ce signal contenu dans le flux perceptif permettrait d’expliquer différentes variations perceptives liées aux contraintes de nos actions. Finalement, cette thèse défend une approche unifiée de la perception, selon laquelle la limite entre les concepts d’action et de perception devrait être repensée afin de rendre compte de la nature sensorimotrice de nos connaissances. / Visual perception of space is mainly known as depending upon one’s vision capacity. However, research about the influence of non-visual variables seems to indicate that the body also plays an important role in visual perception. In this Ph.D. thesis, we support the idea that distance perception has to be studied as a function of the sensorimotor interactions between the individuals and their environment. In the first study, we showed that action constraints in the environment such as energetic cost could influence allocentric distance perception. In the second study, we focused on the role of the sensorimotor integration process in distance perception tasks, as well as the influence of non-visual variables such as audition and motor activity. In the third study, we investigated the phenomenology of perception, and more precisely in the relative fluency of motor activity. In other words, we focused here on the metacognitive feedback that emerges from the quality of the interactions with the environment. We highlighted that such a metacognitive signal could explain the influence of action constraints on distance perception. Finally, this work provides strong supports to the idea of an integrative approach of perception according to which the theoretical boundary between perception and action is questioned by the sensorimotor nature of our knowledge.