We are warming up for our exhibit of Trusted Autonomous Vehicles at the Royal Society Summer Science exhibition from 1st to 7th of July in London. At an Open Day at University of Leicester we trialed our activities - including a demo of a self-driving car on a 3m x 5m track.
Another technology we showcased was an AI-based object detection and recognition. We showed this basic AI component to explain how the artificial intelligence inside an autonomous vehicle is not a Hollywood-style, sentient being but rather a non-trivial piece of software composed on top of basic AI components.
Our third activity for the day was a driving simulator where the driver takes the role of an AI that is then thoroughly tested. Not many drivers that day were awarded a positive score. If these drivers were AIs, we wouldn't let them onto our roads.
Of course with the big event at the Royal Society being three weeks away not everything worked as planned. While some of us seeked help in worshiping the god of self-driving cars...
... others had to fix the bugs they introduced the night before into the software. A rather interesting bug originated from a safety feature that led to a deadlock: I had programmed the pedestrian to not move when the car is immediately in front of it (otherwise the pedestrian might run into the car). This feature of immobility led to a problem when the car had approached, had stopped for, and now was waiting for the pedestrian to leave the road: the pedestrian detected the car and didn't move back.
Some links:
The self-driving car demo illustrated some of the sensors used by autonomous vehicles. We projected the camera view of the car and a visualization of the output of a laser-scanner onto a large screen. In addition, the red buzzer shown on the right (below) controlled a pedestrian to enter the road to demonstrate an emergency brake safety system already in use in many vehicles today.
Another technology we showcased was an AI-based object detection and recognition. We showed this basic AI component to explain how the artificial intelligence inside an autonomous vehicle is not a Hollywood-style, sentient being but rather a non-trivial piece of software composed on top of basic AI components.
Our third activity for the day was a driving simulator where the driver takes the role of an AI that is then thoroughly tested. Not many drivers that day were awarded a positive score. If these drivers were AIs, we wouldn't let them onto our roads.
Of course with the big event at the Royal Society being three weeks away not everything worked as planned. While some of us seeked help in worshiping the god of self-driving cars...
... others had to fix the bugs they introduced the night before into the software. A rather interesting bug originated from a safety feature that led to a deadlock: I had programmed the pedestrian to not move when the car is immediately in front of it (otherwise the pedestrian might run into the car). This feature of immobility led to a problem when the car had approached, had stopped for, and now was waiting for the pedestrian to leave the road: the pedestrian detected the car and didn't move back.
Some links:
- Royal Society Summer Science Exhibition July 1-7
- The website of our DriverLeics team and exhibit
- Two exhibits the University of Leicester will present in London
Comments
Post a Comment