Towards Realistic Walk Path Simulation in Automotive Assembly Lines: A Probabilistic Approach

Abstract

In the last years, automotive production planning has become increasingly time-consuming, since the trend towards mass-customization is currently leading to numerous possible assembly task-sequences within one assembly station. As a consequence of this development, walk paths for assembly operators in a manual assembly line may vary wildly from car to car. An increasing variety of routes, coupled with high requirements for efficiency and working conditions entails a growing demand for realistic walk path planning methods. In practice, walk paths are either planned with pen-and-paper methods or simulated using deterministic motion planning algorithms calculating a two-dimensional trajectory of the worker's Center of Mass. Both methods, however, do not consider gait and its influence on the actual walk path. Furthermore, by applying deterministic simulation approaches, the probabilistic nature of human motion is neglected. As a consequence, actual walk paths can significantly deviate from their corresponding plans. In order to overcome these limitations, this paper presents a two-dimensional motion planner incorporating fine grained information on human gait gathered from 600 000 samples of a probabilistic motion model. Those data points are drawn from a multivariate Gaussian Mixture Model based on real human motion capture data, thus guaranteeing natural human motions. By combining a global motion planner with this motion model, on the one hand, a realistic and collision-free trajectory can be computed in real-time. On the other hand, this novel probabilistic approach contributes to a better prediction quality of planning models by enabling production planning departments not only to calculate one valid walk path but to simulate all possible variants.