Theoretical mechanical benefits of biarticular muscles include reduced displacements and force potentiating shifts in linear velocities during multi-joint coupled motions. A cadaveric model was developed to compute muscle kinematics of biceps femoris (BFL and BFS) during four classes of coupled knee and hip joint motion, as well as running and walking gait (Six subjects, Vicon Motion Analysis). The examples of the classes of motion were: KEHE-jump (knee extension and hip extension), KFHF-tuck (knee flexion and hip flexion), KFHE-kick (knee flexion and hip extension), and KEHF-paw (knee extension and hip flexion). BFL peak and mean velocity shifts relative to BFS were seen in all four coupling classes (p<0.05) and the majority of the gait subclasses (p<0.05). Muscle displacements were larger in BFL for both KFHE-paw and KEHF-kick (p<0.05), smaller in KFHF-tuck (p<0.05), but not significantly different in KEHE-jump or during most of the running gait subclasses, except for during KFHE-late mid stance and KEHF-mid swing, where they were larger for BFL (p<0.05). The mechanical benefits associated with BFL velocity shift relative to BFs were identified in KFHF, KEHF motions, and certain subclasses of gait. In contrast, there were potential mechanical detriments due to velocity shift relative to BFs in the KEHE-jump, KFHE-paw, and the majority of KEHE and KFHE subclasses in both gait cycles. The possible mechanical benefits associated with displacement conservation of BFL relative to BFs would be realized in KFHF-tuck jump, but not during KEHE-jump and the gait cycle subclasses. The findings of this study reveal both mechanical benefits and detriments of biarticular muscles, and have immediate implications for neural control of biarticular muscles during movement.

Theoretical mechanical benefits of biarticular muscles include reduced displacements and force potentiating shifts in linear velocities during multi-joint coupled motions. A cadaveric model was developed to compute muscle kinematics of biceps femoris (BFL and BFS) during four classes of coupled knee and hip joint motion, as well as running and walking gait (Six subjects, Vicon Motion Analysis). The examples of the classes of motion were: KEHE-jump (knee extension and hip extension), KFHF-tuck (knee flexion and hip flexion), KFHE-kick (knee flexion and hip extension), and KEHF-paw (knee extension and hip flexion). BFL peak and mean velocity shifts relative to BFS were seen in all four coupling classes (p<0.05) and the majority of the gait subclasses (p<0.05). Muscle displacements were larger in BFL for both KFHE-paw and KEHF-kick (p<0.05), smaller in KFHF-tuck (p<0.05), but not significantly different in KEHE-jump or during most of the running gait subclasses, except for during KFHE-late mid stance and KEHF-mid swing, where they were larger for BFL (p<0.05). The mechanical benefits associated with BFL velocity shift relative to BFs were identified in KFHF, KEHF motions, and certain subclasses of gait. In contrast, there were potential mechanical detriments due to velocity shift relative to BFs in the KEHE-jump, KFHE-paw, and the majority of KEHE and KFHE subclasses in both gait cycles. The possible mechanical benefits associated with displacement conservation of BFL relative to BFs would be realized in KFHF-tuck jump, but not during KEHE-jump and the gait cycle subclasses. The findings of this study reveal both mechanical benefits and detriments of biarticular muscles, and have immediate implications for neural control of biarticular muscles during movement.

Emergency first responders are the great heroes of our day, having to routinely risk their lives for the safety of others. Developing robotic technologies to aid in such emergencies could greatly reduce the risk these individuals must take, even going so far as to eliminate the need to risk one life for another. In this role, humanoid robots are a strong candidate, being able to take advantage of both the human engineered environment in which it will likely operate, but also make use of human engineered tools and equipment as it deals with a disaster relief effort. The work presented here aims to lessen the hurdles that stand in the way through the research and development of new humanoid robot technologies. To be successful in the role of an emergency first responder requires a fantastic array of skills. One of the most fundamental is the ability to just get to the scene. Unfortunately, it is at this level that humanoid robots currently struggle. This research focuses on the complementary development of physical hardware, digital controllers, and trajectory planning necessary to achieve the research goals of improving the locomotion capabilities of a humanoid robot. To improve the physical performance capabilities of the robot, this research will first focus on the interaction between the hip and knee actuators. It is shown that much like the human body, a biped greatly benefits from the use of biarticular actuation. Improvements in efficiency as much as 30% are possible by simply interconnecting the hip roll and knee pitch joints. Balancing and walking controllers are designed to take advantage of the new hardware capabilities and expand the terrain capabilities of bipedal walking robots to uneven and non-stationary ground. A hybrid position/force control based balancing controller stabilizes the robot's COM regardless of the terrain underfoot. In particular two feedback mechanisms are shown to greatly improve the stability of bipedal systems in response to unmodelled dynamics. The hybrid position/force approach is shown through experiments to greatly extend humanoid capabilities to many types of terrain. With robust balancing ensured, walking trajectories are defined using an improved linear inverted pendulum model that incorporates the swing leg dynamics. The proposed method is shown to significantly reduce the control authority (by 50%) required for satisfactory trajectory following. Three parameters are identified which provide for quick manual or numerical solutions to be found to the trajectory problem. The walking and balance controller were operated on four different terrains successfully, strewn plywood, gravel, and high pile synthetic grass. Furthermore, SAFFiR is believed to be the first bipedal robot to ever walk on sand. The hardware enabled force control architecture was very effective at modulating ground reaction torques no matter the ground conditions. This in combination with highly accurate state estimation provided a very stable balance controller on top of which successful walking was demonstrated. / Ph. D.

Ce travail de thèse s’inscrit dans le cadre de la simulation par mannequins numériques de l’homme (en anglais Digital Human models ou DHM) en vue d’applications ergonomiques. Plus précisément, il s’intéresse à la modélisation du système musculosquelettique et des efforts musculaires développés au cours du mouvement.Dans un première partie est présenté le développement numérique de modèles musculosquelettiques du membre inférieur et supérieur sous l’environnement Matlab. Ces modèles ont été évalués par examen des bras de levier musculaires pour la géométrie, et par comparaison avec des mesures expérimentales de couples articulaires maximaux (FMV) pour les efforts prédits.Dans une seconde partie, le modèle musculosquelettique du membre inférieur a été utilisé pour évaluer le rôle des muscles bi-articulaires au cours de la marche. Pour cela deux méthodes de calcul des efforts musculaires ont été utilisées, permettant de mettre en évidence les particularités fonctionnelles de ces muscles. Les résultats indiquent notamment que la prise en compte des muscles bi-articulaires est nécessaire à une estimation correcte des efforts musculaires et des efforts de contact articulaires, potentielles sources d’inconfort.Puis ont été étudiés les capacités d’effort maximale en flexion-extension isométrique du coude. L’expérimentation a été effectuée au sein de l’INRETS de Bron et comprenait 9 sujets volontaires masculins. Cette étude a permis d’évaluer le modèle numérique développé en termes de prédiction de capacités d’effort, ainsi que d’ouvrir la voie à des méthodes de scaling permettant d’ajuster un modèle musculosquelettique au sujet étudié. / This work deals with the use of Digital Human models (DHM) for ergonomic applications. More precisely, it focuses on the modelling of the musculoskeletal system and muscular forces developed during movement.In the first part, the development of numerical musculoskeletal models of the lower and upper limbs under the Matlab environment is presented. These models were evaluated by examination of the muscles’ lever arms (for geometry) and by experimental measurements of maximum voluntary force (MVF) (for predicted muscle forces).In a second part, the lower limb muscular model was used to evaluate the role of biarticular muscles during gait. To achieve this, two methods were used, allowing to highlight the functional specificities of these muscles. The results indicate in particular that taking into account biarticular muscles is necessary to a correct estimation of muscle forces and joint contact forces, which are potential discomfort sources.Finally, the maximal force capacity of isometric elbow flexion-extension has been studied. The goal was to set up scaling methods allowing to fit the muscular model to the studied subject’s capacities. The experiment took place at INRETS, and implicated 9 male voluntary subjects. This study allowed to evaluate the model in terms of maximal force capacity prediction, and openend the path to new scaling methods for musculoskeletal models.

In this dissertation, we describe the development of a humanoid bipedal robot that fully physically models the human walking system, including the biomechanics of the leg, the sensory feedback pathways available in the body, and the neural structure of the central pattern generator (CPG). Using two different models of the CPG, we explore several issues in the neurobiology and robotics literature, including the role of reflexes in locomotion, the role of load reception and positive force feedback in generating the gait, and the degree to which central or peripheral control plays in human walking. We show that the walking pattern can be generated by a combination of a half-center CPG and reflex interactions phase modulated by the CPG, and that load receptors in the muscles can play a substantial role in generating the gait, using positive force feedback. We compare the gait of the robot to human subjects and show that this architecture produces human-like stepping. Varying the degree of direct central control of lower limb muscles by the CPG, we show that the most human-like gait is generated with a relatively weak central control signal, which modulates reflex responses that generate most of the muscle activation. These results allow us to conceive of locomotion as a series of nested loops, with a central CPG or rhythm generator modulating lower level reflex interactions, while higher centers modulate the CPG. Since locomotion is a primary mechanism by which animals interact with the world, this research is relevant to artificial intelligence researchers. Recent understanding of cognition holds that minds are embodied, situated relative to a set of goals, and exist in a feedback loop of interaction with the environment. In our robot, we model the dynamics of the body, the neural architecture and the sensory feedback channels in a complete dynamical feedback loop, and show that the robot entrains to the the natural dynamics of the world. We propose the concept of nested loops with descending phase modulation as a conceptual paradigm for a more general understanding of nervous system organization.