Rapidly locating and accurately tracking the center of mass using statically equivalent serial chains

This dissertation presents a center of mass (CoM) estimation technique that uses the statically equivalent serial chain (SESC). A SESC is a representation of any multilink branched chain whose end-effector locates the CoM. Identifying the center of mass location provides a significant aid in controlling the balance of humanoid robots. Additionally, in humans this location is an essential parameter in postural control and is critical in assessing rehabilitation. Anthropometric tables have been complied for this identification but their accuracy is readily questioned. The method starts with an experimental phase involving a force plate and a motion capture system (MoCap) to construct a model to predict the CoM location. Subsequent motion of the subject updates the CoM model based on MoCap information without need of a force plate, overcoming disadvantages of some other CoM estimation methods. The node-based SESC model is developed to best integrate with Kinect and the likely MoCap systems that will be developed in the near future. The results show that the SESC methodology allows rapid and accurate real time estimation of the CoM. The transfer-ability of the SESC parameters among subjects with similar body structure using the donor model is also presented. The donor model allow subjects to perform fewer postures in the experimental phase to generate a SESC. The donor model facilitates the identification of a SESC for subjects with limited mobility. This work includes the CoM estimation for human subjects using low-end and high-end MoCap and force plate sensors. The high-end and low-end MoCap and force plate sensors are used for cross validation and to show that the method is applicable to both systems. Additionally, the presence of a static body in the workspace (a walker or chair, for example) to create stability in test subjects is presented. The introduction of the static body aids in balance and stability and adds more postures to agile subjects. Furthermore, the effect to the SESC parameters by the modification of the node-based SESC technique to allow an arbitrary location of the CoM associated with the torso is presented. The addition of a moving frame to any specific node accounts for CoMs that do not lie along the line connecting successive nodes. Finally, the accuracy of the vertical component of the CoM is also considered