Unified Effort Measure for Natural Posturing of Digital Humans

Abstract

Digital Human Models (DHMs) are extensively used in industrial ergonomic simulations to predict, understand, and analyse human behavior. However, most existing DHMs are built upon principles developed for robotic manipulators, where physics and geometric principles characterize the stability and control of postures. However, these approaches do not capture the inherent human aspects of motion, viz. sensory feedback, perceptual judgment, and adaptive behavior that help avoid unstable or awkward postures during task execution. Consequently, current DHMs often produce oversimplified and unrealistic human posture and motion. This thesis addresses these limitations by presenting a sensing-based adaptive framework that unifies stability, motion, and effort considerations for more realistic and responsive digital human performance simulations in ergonomic assessments.

The thesis introduces a novel sensing-based scheme for postural stability, which uses support reaction forces estimated through the so-called Global Mean Value coordinates. The method is shown to be equivalent to the conventional physics-based assessment of stability. To overcome the limitations of the conventional binary notion of stability, two real valued, perception-based metrics, namely Subjective Stability Measure and Subjective Discomfort Measure are proposed. The strong correlation observed between comfort and stability, suggest that the traditional geometric check for stability is redundant within the proposed sensing-based model. For the first time in literature, it is demonstrated through simulation that the empirically observed Functional Stability Region could emerge naturally through subjective thresholds on local pressure in the support-interface.

Taking inspiration from the well-known psychophysical framework of perception-cognition-action, the thesis proposes a posture evolution methodology wherein the human movement is modelled as a continuous sequence of controlled postures guided by sensory input, perceptual assessments, and task demand; this significantly departs from the conventional posture prediction framework in DHMs which aim at obtaining a realistic final posture alone. For capturing the strategic recruitment of body segments of human, a novel joint-space partitioning scheme is proposed which groups the joints based on their roles in a task. Additionally, the method simulates natural postural transitions, ensuring that each intermediate posture is stable, comfortable, as well as biomechanically feasible.

The above features enabled the development of methods to estimate the functional workspace and hand trajectories for reach tasks. To enhance the utility of the conventional notion of a workspace redefined as a fixed region of kinematic reachability, a task-specific, functional space constrained by biomechanical feasibility is proposed in this work. The proposed Unified Effort Measure (UEM) combines the intrinsic joint effort measure with the extrinsic support discomfort measure to quantify the postural cost of reaching points in a grid-sampled space. Tasks like sequential/cyclic reach and pick-and-place with varying loads on the hand, commonly seen in mechanical assembly tasks, are simulated in the UEM rendered space. Hand paths are then optimized by minimizing cumulative effort, revealing how task location, external forces, and other constraints shape natural motion paths. It demonstrated that reachability alone is insufficient for task execution as the functional workspace is a subset of the reachable workspace. The hand trajectories are shown to arise naturally from the intent to minimize the total postural effort while dynamically adapting to the external loads.

In summary, this thesis proposes a human-centred DHM framework that integrates sensory feedback, effort minimization, and task specific constraints to realistically simulate posture and motion, laying the foundation for adaptive, next-generation Digital Human Models.