Virtual Lidar Sensor Intensity Data Modeling for Autonomous Driving Simulators

Abstract

Autonomous driving simulators are an effective tool for developing autonomous driving algorithms so that they are widely used in research and development. However, the similarity of the virtual model to reality is closely related to the validity of the simulation results. Therefore, analyzing the characteristics of real sensors is necessary for mathematical modeling of virtual sensors in autonomous driving simulators. This paper presents a virtual lidar that has a high fidelity operating similarly to reality in the sensor modeling. The intensity variable factors which represents the strength of the received signal relative to the transmitted signal are effectively used for improving the fidelity of virtual lidar with a low computing power. The proposed virtual lidar is implemented in an autonomous driving simulator to show its feasibility by comparing with the existing virtual lidar. Specifically, the paper focuses on modeling intensity data with an aim to exhibit trends similar to the intensity measurement results of real lidar, compared to conventional modeling methods. The proposed virtual lidar is implemented in an autonomous driving simulator to demonstrate its feasibility by comparison with existing virtual lidar. As a result, there was an 89.21 percent improvement in the average intensity within the region of interest compared to the conventional modeling method.

Keywords

• Driving simulator
• fidelity
• lidar sensor model
• intensity
• sim2real