Printer friendly version
Car Steered with Eyes: Demonstration by Computer Scientists of Freie Universität at Former Tempelhof Airport
23 April 2010
Freie Universitaet Berlin
"Keep your eyes on the road!" - Today scientists at Freie Universität working under the computer science professor Raúl Rojas gave a completely new meaning to this standard rule for drivers: Using software they developed, they can steer a car with their eyes.
Today on the site of the former Berlin Tempelhof Airport, the head of the project, Raul Rojas, and his team from the Artificial Intelligence Group demonstrated how they can steer the vehicle that is equipped with complex electronics just by eye. More than 60 journalists from around the world were there to watch.
Information about the Software: EyeDriver
The eyeDriver software is a prototype application for steering the research vehicle Spirit of Berlin using eye movements. The software was designed by computer scientists at Freie Universität Berlin in collaboration with the company, SMI (SensoMotoric Instruments). The eye movements of the driver are collected and converted into control signals for the steering wheel. The speed is controlled separately and is not included in eyeDriver. The software shows that you can drive a vehicle alone with eye movements.
The HED4 solution by SMI is used for detecting and tracking the eye movements. It is a converted bicycle helmet equipped with two cameras and an infrared LED, as well as a laptop computer with special software. One of the cameras is pointed to the front in the same direction as the person wearing the helmet (scene camera), while the other camera films one eye of the wearer (eye camera). The infrared light supports the eye camera and is pointed to the eye under observation. A transparent mirror that reflects only the infrared light is used to allow a reasonable viewing angle for the eye camera, without limiting the wearer's ability to see. After a brief calibration the software on the laptop of the HED4 is not only able to capture the position of the pupil in the eye camera, but can also calculate the position in the scene camera that the wearer is looking at. These coordinates in the image of the scene camera (viewing position) are transmitted via an ordinary LAN to the onboard computer in the Spirit of Berlin. The eyeDriver software in the onboard computer in the Spirit of Berlin receives the viewing positions at regular intervals over the LAN in the vehicle and uses it to control the steering wheel. The driver can choose between two modes: "free ride" and "routing."
In the "free ride" mode the viewing positions are linked directly with the steering wheel motor. That means that the x-coordinates of the viewing position are used to calculate the desired position of the steering wheel. The further the driver looks to the left or right, the further the steering wheel is turned in that direction. The speed of the vehicle is set in advance and kept constant, as long as the position of the gaze is recognized. In case it is not possible to detect which direction the driver is looking in, for example, if the driver's eyes are closed, the vehicle brakes automatically.
In the "routing" mode, the Spirit of Berlin steers autonomously most of the time. Only where there is a fork in the road, or an intersection, the car stops and asks the driver to select the next route. This requires the wearer of the helmet to look to the left or right for three seconds. If the driver's gaze lingers long enough in one direction, the eyeDriver software confirms acoustically that the choice has been accepted. The decision is communicated to the planner in the vehicle. Then the artificial intelligence in the Spirit of Berlin can plan the route accordingly and continue to run independently.
The Autonomous Vehicle Project
Prof. Dr. Raúl Rojas is a professor of artificial intelligence at the Institute of Computer Science at Freie Universität Berlin. He gained international success with his soccer robots, the "FU-Fighters." They were world champions twice in the small-size league. Since 2006 Prof. Rojas and his team have been designing technologies related to autonomous vehicles. As part of this project, they developed the research vehicle Spirit of Berlin, making it to the semifinals in the DARPA Urban Challenge in California in 2007.
In the fall of 2009, in the innovative vehicle steering series, based on the test vehicle, the computer scientists Tinosch Ganjineh and Miao Wang developed iDriver, with which it is possible to steer the research car using an iPhone. This series is now complemented with the EyeDriver software. It was developed primarily by Miao Wang and David Latotzky in cooperation with the firm SMI. These two developments are simply sub-projects. The core of the research continues to be the autonomous driving.
The AutoNOMOS Project
Since November 2010 the team around Prof. Rojas has been working on the further development of autonomous or semi-autonomous cars in the AutoNOMOS project headed by Tinosch Ganjineh. The project is being funded by the German Ministry of Education and Research (BMBF) in its ForMaT (Forschung für den Markt im Team) program. The funding is for two years. The project will make a significant contribution to the development of accident-free, efficient, and environmentally friendly mobility. AutoNOMOS is a modular system for the operation of autonomous or semi-autonomous cars. Using AutoNomos, it will be possible to detect impending dangers on roads, highways, and crossings (lane change, traffic jams, rights of way) at an early stage and accidents can be prevented. Once the technology is ready, it will be introduced at first on private property and, finally, in public traffic.
The Spirit of Berlin
The Spirit of Berlin is an autonomous vehicle that was designed and built by the Artificial Intelligence Group at Freie Universität Berlin as of 2007. It is a car that can be driven without a driver. A conventional car (Dodge Grand Caravan, 2000) was equipped with sensors, computers, and actuators. The sensors collect information about the immediate environment. Using this information, the software on the computers selects what to do. The resulting action is then implemented with the help of mechanical actuators.