Generic motion models based on quadrupedal data

Motion capturing has become a standard technique in computer graphics and biomechanics. Animal locomotion has been recorded in different environments and this has been successfully used in animal biomechanical experiments. There is an increasing interest to acquire and analyse animal motion data. Specifically, spinal quadrupedal locomotion is currently an area of focused interest, and the factors influencing this motion such as load and force transmission from the limbs to the head, neck, and back has not yet been evaluated. A comparative method investigating different spinal anatomy and movement offers the possibility to develop a more complex and advanced general theoretical model providing a more unified theory of locomotory mechanisms and control. The long-term goal of the proposed project is to identify generic data-driven and biomechanics-based methods that can be used for capturing, analyzing, up to synthesizing mammalian quadrupedal motions. There are many inter- correlations and loops in the relations of these tasks, some of which are of primary interest in the area of computer graphics and others in biomechanics research and veterinary medicine. Based on the hypothesis that spine flexibility is of essential importance for understanding quadrupedal motions, the development of a generic quadrupedal motion model parameterized by spine flexibility will be the central goal for the overall project. The investigation of overall movement in relation to spinal curvature and flexibility by modelling different curvatures and flexibilities can quantify the active and passive contribution of the quadrupedal spine. Changing curvature and flexibility parameters of the spine will therefore influence the motion pattern; and the development of such a biomechanical model is necessary to fully characterize static and dynamic stability. Once this basic relationship between spinal shape and function has been identified, dynamic simulation can be used to quantify the forces and moments on spines of different shape in walk, trot and canter. A relational database will be established that contains motion capture data from different mammalian quadrupeds. Data will be recorded using optical motion capture, accelerometers and video. The data will then be freely accessible. The accuracy of models determines their usefulness for any application. Accurate representation of individual responses to force is necessary when attempting to develop models that are realistically reproducing the mechanisms associated with a specific response. Optimization of such models requires a quantitative measure of their accuracy. Comparing the measured data with the accuracy obtained from simulating the motion using the generic model developed in this study, as well as with the accuracy of using alternative models will be done. Finally, this will allow an evaluation of the confidence with which certain levels of accuracy can be achieved in defined conditions for the generic model.

The joint efforts of the computer science work-group in Bonn and animal biomechanics work- group in Vienna have produced a successful collaboration establishing a system to obtain a novel collection of animal movement data. Furthermore, the collaboration succeeded to study locomotion and movement of quadrupeds both inside a laboratory setting as well as outside of the laboratory environment. This has widened the possibilities as previous investigations were restricted to a particular space and because of this also limited in the gait and locomotion events that they were able to record, as not all of them are displayed in a laboratory setting. Moving outside of the laboratory setting made it possible to register more freely and more naturally moving animals and also animals not amenable to be introduced to the laboratory. In the present project, several techniques were used to record these various movement patterns and locomotion events. These included inertial sensors to detect three-dimensional movement patterns and in parallel continuous video recordings were collected to gather additional information about the animals (e.g. facial expressions), as well as to bring us a step closer to use video data analysis for movement documentation instead of sensor based technologies. The research project also employed the documentation of muscle activities to the movement/locomotion documentation leading to a wider scope of the project, including the questions of general balance and muscle control mechanisms of movement in these quadrupeds. The results of the project lead to a quantification of similarities and dissimilarities of movement patterns, with the focus placed on the back movement of the animals, based on vertebral column movement and flexibility. The findings showed that, while there are individual differences as well as differences between the quadrupedal species investigated, the overall pattern was similar in the sheep and horses investigated, both members of the ungulates. Uniting the strength of the field of computer science and animal biomechanics made it possible to investigate research questions in a transdisciplinary approach not accessible to either of the research groups alone. On a personal development level, the project introduced the students to working in an international and even more importantly a transdisciplinary setting, which will help be of great importance also for the future generations in basic science. The successful bridging of the discipline specific differences in approach and in terminology as well as the effort necessary to reach a common understanding of the project details have been very well illustrated. This project has established an even stronger collaboration between these science disciplines and between these research groups, and we expect that future projects based on this cooperation will have a considerable chance for funding success of projects developed together as well as publishing success of the results.