ABSTRACT

Although the degree of automation is increasing in manufacturing industries, many assembly operations are performed manually. To avoid injuries and to reach sustainable production of high quality, comfortable environments for the operators are vital. Poor station layouts, poor product designs or badly chosen assembly sequences are common sources leading to unfavorable poses and motions. To keep costs low, preventive actions should be taken early in a project, raising the need for feasibility and ergonomics studies in virtual environments long before physical prototypes are available.

Today, in the automotive industries, such studies are conducted to some extent. The full potential, however, is far from reached due to limited software support in terms of capability for realistic pose prediction, motion generation and collision avoidance. As a consequence, ergonomics studies are time consuming and are mostly done for static poses, not for full assembly motions. Furthermore, these ergonomic studies, even though performed by a small group of highly specialized simulation engineers, show low reproducibility within the group.

Effective simulation of manual assembly operations considering ergonomic load and clearance demands requires detailed modeling of human body kinematics and motions as well as a fast and stable inverse kinematics solver. The focus in this paper is to evaluate and compare the performance of different optimization 286algorithms. This is done by letting them find an optimum of the ergonomic function while kinematic constraints are fulfilled.

The manikin used in this work has 162 degrees of freedom and uses an exterior root. To describe operations and facilitate motion generation, the manikin is equipped with coordinate frames attached to end-effectors like hands and feet. The inverse kinematic problem is to find joint values such that the position and orientation of hands and feet matches certain target frames during an assembly motion. This inverse problem leads to an underdetermined system of equations since the number of joints exceeds the end-effectors' constraints. Due to this redundancy there exist a set of solutions, allowing us to pick a solution that maximizes a scalar valued comfort function.

Finding an optimum for the non-linear comfort function can be done with different algorithms, but the choice is not obvious. Therefore, in this paper we will implement, evaluate and compare the algorithms Interior Point and the so called Resolved Motion Rate. This will lead to a better understanding of their pros and cons, and give us the possibility to further improve the current way of generating motions for this type of digital human models.

The methods will be tested on a large number of random motions and on a set of challenging assembly operations taken from the automotive industry.