3D Érzékelés és Mobilrobotika Kutatócsoport

Mobil robotika

Mobile robotics has been a rapidly expanding area of research in the last two decades. With the recent developments of artificial intelligence the focus of research has shifted towards autonomous vehicles. With a vast area of applications ranging from cleaning robots, robots aiding the elderly of the disabled to driverless cars, this field of research is certain to shape the future of men and machines. The research in our lab focuses on the perception problems arising with mobile robots. Since these robots often operate autonomously, they have to perceive and react to their environment intelligently. We have a remote control car system used for education purposes, aiming to introduce students to the problems of teleoperation and remote sensing. We also have a racing car which we are converting into an autonomous vehicle capable of beating a visual obstacle course.

Visual Obstacle Course for Mobile Robots - Autonomous robots and vehicles are a significant area of research at our lab. Since autonomous decision making requires plenty of classical vision or similar perception tasks, our lab focuses on robot vision mainly. Our lab has a trophy Flux Buggy racing car and several visual sensors, such as a Bumblebee stereo camera and a Kinect sensor, which may be used to build an autonomous mobile robot system.

Driver Assistance Systems - The spread of the driver assistance systems and the growing importance of their vision component open new possibilities for constructing such database, which describe the environment online in deep details. The onboard vision system of a state of the art vehicle delivers much more information than we need for its basic tasks. By use of additional evaluation of the visual information, and with the installation of extension elements and communication tools we achieve a cheap data source for a couple of new applications.

3D Visual SLAM - One important aspect of autonomous intelligent robots is their ability to navigate in unknown environments. These robots are able to perform simultaneousl localization and mapping (SLAM), meaning they construct a map of their environment as they explore it, while determining their position at the same time. In this project, our aim is to implement a SLAM algorithm using the robot's internal sensors and a 3D camera, allowing for robust navigation in complex settings.

Traffic Sign Detection for Autonomous Cars - Visual perception of the environment is an essential task for any autonomous vehicle. In the case of everyday traffic recognising traffic signs is a fundamental part of this problem. While there are increasingly precise online databases available for traffic rules, still, combining the data received from these with real time perception is recommended in order to avoid errors caused by inaccuracies, slow updates or loss of connection.

Collision Detection and Avoidance - Perhaps the most important aspect of the science of autonomous vehicles is the detection and avoidance of accidents or other dangerous scenarios. In fact, frontal collision detection and automatic braking systems have been in the market for quite a few years. In many cases, however, braking is not enough to avoid accidents, therefore course adjustment is also needed. This introduces several problems, namely the assessment of safe avoidance paths. In our research we tackle the problem of frontal collision detection and the finding of safe avoidance paths using a forward-facing camera.