Getting your Trinity Audio player ready...
|
A team of researchers from The Ohio State University has created innovative software designed to assist in the advancement, assessment, and demonstration of autonomous vehicles, ensuring safety. Known as Vehicle-in-Virtual-Environment (VVE) method, this software enables the testing of driverless cars within a completely secure environment.
Bilin Aksun-Guvenc, a Professor of Mechanical and Aerospace Engineering at Ohio State and Co-author of the study, explained the capabilities of the VVE method. “Through our software, we can create the illusion that the vehicle is driving on real roads while it operates within a secure and extensive testing environment. This capability saves time and money and eliminates the risk of fatal traffic accidents.”
The research demonstrated that immersing autonomous vehicles in a virtual environment allows them to learn how to prevent car collisions, enhance pedestrian safety, and respond effectively to uncommon or extreme traffic situations.
Aksun emphasised that despite the increasing prevalence of autonomous driving technologies on the roads, the study highlights the need to carefully examine the testing methods employed, considering the significant number of accidents attributed to these systems.
“Our research endeavours are focused on achieving that goal, where we can have complete confidence in the safety of all road vehicles. Trust is vital for our future,” said Aksun.
According to Aksun-Guvenc, existing methods for showcasing autonomous vehicle capabilities involve testing software and technology through simulations and real-world road trials. However, this approach involves other road users as involuntary participants in these experiments, introducing risks that can lead to high costs, inefficiencies, and potential hazards for drivers and pedestrians.
To address the shortcomings of inadequate evaluations, the researchers in this study introduced a novel approach. They substituted the real-time data from high-resolution sensors in an actual vehicle with simulated data, effectively connecting the vehicle’s controls to a highly realistic 3D environment. This method can be likened to equipping the vehicle with virtual reality glasses or a VR headset.
By inputting the simulated data into the autonomous driving system’s computers and synchronising the actual movements of the car with the virtual simulations, the researchers successfully demonstrated that the vehicle responds as if the virtual environment were its genuine surroundings, all in real time.
However, their software’s exceptional power lies in their virtual environment’s remarkable flexibility, as highlighted by Levent Guvenc, co-author of the study and co-director of the Automated Driving Lab. The model can be readily adjusted to suit various scenarios by replacing actual sensory inputs with virtual ones.
“The VVE method allows for calibration that preserves real-world characteristics while simulating rare events in the virtual environment,” explained Guvenc.
The software can effortlessly simulate various traffic scenarios, from extreme situations like someone unexpectedly jumping in front of a vehicle to mundane occurrences like pedestrians waiting at a crosswalk.
Using Bluetooth technology, the software establishes communication between a pedestrian’s mobile phone and the test vehicle. In a demonstration, the virtual darting of a pedestrian across the road, at a safe distance from the vehicle was interpreted by the car as if the person was darting directly in front of it.
“The remarkable aspect of this method is that road users can share the same environment simultaneously without physically occupying the same location at all,” explained Guvenc.
While creating these highly realistic environments may require time, Guvenc mentioned that their team has successfully addressed the technological challenge of synchronising different environments for real-time simulations.
The team has also taken steps to protect its innovation by filing a patent for the technology. Afterwards, Guvenc expressed his desire to see its integration into traffic guidelines established by organisations such as The National Highway Traffic Safety Administration.
Guvenc believes that this technology has the potential to become a standard practice within the industry in the next five or ten years. As a result, the team is actively focusing on developing more applications for its utilisation in the future.