Event-Based Visual-Inertial Odometry(NPO-52048-1)

Estimating the position of a robot (localization) is an important precursor to enabling a robot to navigate autonomously. Traditional camera-based localization methods fail when the robot is moving fast (motion blur) or when the robot is in very dark or bright environments (over/undersaturated pixels). An event camera is a new type of imaging sensor that reports changes in pixel brightness per pixel independently and asynchronously with microsecond resolution and overcomes both of these problems. However, the asynchronous nature of the data is a challenge for traditional "synchronous" pipelines that receive a full image at a time. This software offers a couple of approaches to processing this asynchronous event stream to estimate the position of an event camera.