Digital humans in CAD systems can facilitate product and workspace designs through timely virtual ergonomic analyses. Currently, digital humans lack the capability of faithfully simulating human motions and motor capabilities. Existing methods are limited in that they do not provide: (1) planning of different types of motions on a unified framework, (2) consideration of alternative movement techniques, and (3) automatic adoption of new motions.; To overcome these limitations, this research proposes a new approach for human motion planning termed Memory-based Motion Simulation (MBMS). In this context, motion planning utilizes joint angle templates stored in memory as the roots for producing new motions while conserving the properties of natural motions.; The MBMS system consists of a motion database, a root motion finder, a movement technique classifier, and a motion modification algorithm. Given an input simulation scenario, the root motion finder searches the database for motions likely to approximate the scenario within specified limits. The selected motions are called root motions.; Root motions may represent a mixture of fundamentally different ‘alternative movement techniques.’ To represent them, a Joint Contribution Vector (JCV) was developed to characterize a motion by quantifying the contributions of each joint motion to its movement goal achievement. Once characterized as JCVs, root motions can be grouped by statistical clustering to reveal alternative movement techniques.; A motion modification algorithm alters root motions to meet given task requirements. The algorithm first identifies a root motion's underlying structure by resolving joint angle trajectories into sequences of motion primitives. This allows parameterization based on amplitude- and time-scaling for generalization of a root motion. An optimization scheme then modifies the root motion's joint motions according to new scenarios while maintaining the original structure, and minimizing changes in the angular velocity profiles and the terminal postures.; Experimental testing was performed for validation. The MBMS was found to accurately predict various seated reaches and whole-body load transfer motions. The robustness of the JCV was demonstrated by its ability to distinguish lifting techniques and identify alternative movement techniques for whole-body motions. A case study demonstrated the utility of MBMS in improving the ergonomics of manual handling tasks.
展开▼
