GPS-AIDED VISUAL-INERTIAL ODOMETRY FOR SELF-DRIVING CARS

Abstract

Automated systems are the way the world is going. As for automation, the concept of self driving cars is under significant hype these days as it has significantly reduced the accidents caused by human error. However, this industry is working on making the ride safer and safer, so the promises made by the developers of autonomous vehicles are simple and appealing. Moreover, after almost many years of hype and headlines, it turns out that the challenges have been much more complicated than anticipated. As for our project, we are working on the state estimation for self-driving cars. Our project aims to improve self driving cars' state estimation by fusing GPS sensors with the Visual Inertial Navigation model (VINS). This will significantly help the vehicle address and alleviate most self driving cars' drift issues. We are using different sensors placed at strategic vantage positions on the car. Our GVINS model gives us the final state sentiments of the vehicle. An Extended Kalman Filter (EKF) is used to predict the final estimate using GPS and VINS. This cutting-edge technology helps significantly improve the accuracy of self-driving cars.