Modeling and Kinematic Optimization of the Human Upper Limb for Collaborative Robotics

The emerging research field developed to optimize the collaboration of human-robot systems for Industry 4.0 gives a central role to the tracking of human motion. Inertial Measurement Units (IMUs) represent a suitable solution to unobtrusively monitor workers in the industrial environment. However, the computation of IMUs orientation usually causes drift problems and affects the kinematics estimate. Moreover, the traditional Euler angles decomposition from the mutual independent orientation of IMUs is affected by mathematical singularities and it does not include joint constraints to avoid violation of physiological motion range. To overcome these limitations, this work aimed at developing a Denavit-Hartenberg upper limb model consistent with standard biomechanical guidelines and an optimization framework for the real-time tracking of human motion. At each time step, the joint variables of the model were estimated minimizing the difference between the modeled segments orientations and those obtained with the sensor fusion. The proposed method was validated with synthetic and real robot data, verifying the influence of a considerable drift on the estimate accuracy. Finally, a comparison between the optimized joint kinematics and the one obtained with traditional methods was made.