Sensor Simulation and Evaluation for Infrastructure-Free Mobile Sensor Carrier Platforms

Abstract

In this paper, we propose a highly accurate way to simulate Time-of-Flight sensors and their explicit characteristics, those enabling sensor evaluations for critical applications, like measurement scenarios, human-machine-collaboration and autonomous driving, as well as a way for estimation of the best-suited sensor for a specific use case. A. Schütz did the development of the simulation, while M. Groneberg was focused on the application of such simulation to logistics. Time-of-Flight Sensors are able to provide robust depth images of dynamic scenes at high framerates, which makes them interesting for movement analysis and logistics use cases and makes them advantageous for autonomous mobile robot navigation. Yet the depth images are disrupted by a number of systematic errors, which, to our knowledge, have not been fully simulated so far. These errors has to be taken into account in early development stages to avoid the destruction of hardware or injuring humans working with autonomous robots. We believe that simulations are suited to develop and to test algorithms for navigation under different environmental conditions to avoid errors, which are caused by misleading depth images. This paper examines the systematic errors of three different Time-of-Flight sensors and presents a physically motivated approach to simulate those sensors. We were able to simulate most of the systematic errors, which were observed in depth images of time-of-flight sensors by using a path tracing algorithm to calculate the influence of light propagation in a physically accurate manner.