Simulation of Industrial Bin Picking: An Application of Laser Range Finder Simulation

Abstract

In bin picking, robots manipulate randomized objects placed in a bin. For that, the objects have to be located before picking. The procedure of localization relies heavily on data from visual sensors, i.e., laser range finders. Development and testing of robot cells with optical localization is time-consuming because realistic sensor data is hardly available. This paper addresses this problem by presenting a framework for simulating robotic bin picking cells. The framework includes a representation of a virtual sensor model for laser range finders, which considers different sources of noise. Well-known ray tracing methods are used to generate synthetic three-dimensional point clouds representing the virtual scene realistically by applying an additive Gaussian Error model. Encouraging results for the simulation of bins filled with gear shafts are presented.