Dynamic vision interface for TurtleBot

Abstract

The goal of this master's thesis is to develop a ROS2 driver able to retrieve and communicate the different data captured by the event-based cameras from iniVation. The implementation includes a capture node which reads the data from a camera and publishes them to their corresponding topics, as well as two other nodes in charge of the visualization of the data. To assess the quality of the driver, we performed a latency analysis and ensured that it was able to transmit the data in a few milliseconds. Following this, we designed as a proof of concept a way to control a TurtleBot based on the output of the camera. Finally, we completely integrated the camera with the TurtleBot. Facing latency issues, we managed to show that the Raspberry Pi on which the TurtleBot is built was not powerful enough to process all the events captured by the cameras. We eventually managed to overcome these problems by limiting the number of events processed by the Raspberry Pi.