Many driving conventions and unwritten rules are highly dependent on eye contacts between drivers such as those of 4-point stop crossing, change lanes, U-turn, and narrow road passing etc. In other words, safe driving does not only be dependent on traffic lights, but also rely on drivers’ cognitive ability with other road users. How would computer-driven cars implement these human-level cognitive capabilities?
Hi everyone, I am following this interesting question to see if there will be some interesting ideas. Driverless cars still have a number of serious limitations. Coming from the computer and internet field, I find this topic intriguing because smart cars are also an example of what we call "Internet of Things" as they are constantly connected and communicating. I think one of the main problems here would be the mixture of both driverless cars and human drivers in the same traffic. If all cars were smart and driverless, they could signal and track one another, and being programmed in a standard way, they would follow similar patterns of behavior. But human drivers are much more complicated and unpredictable. Perhaps we can make an analogy between robots operating in factories vs humans. I think traditionally they tended to be separated for safety reasons and for protecting humans, and only recently after so much work and refinement, they are working together in the same factory lines. Another part of the problem is that human drivers are not perfect either. Dr. Wang is right in the idea that there needs to some type of a negotiation interface that can deal with other drivers and make decisions accordingly. More artificial intelligence might be needed here. Perhaps the logic of these cars can be enhanced with ideas from expert & experienced drivers...
http://www.dailymail.co.uk/sciencetech/article-2739765/Is-Google-s-driverless-car-fair-weather-friend-Vehicle-WON-T-work-heavy-rain-snow-insider-reveals.html
http://www.robotics.org/content-detail.cfm/Industrial-Robotics-Industry-Insights/The-End-of-Separation-Man-and-Robot-as-Collaborative-Coworkers-on-the-Factory-Floor/content_id/4140
This technology works on gap acceptance and headway model between two vehicles . So there is algorithm that solve the problem using GPS technology .Therefore eye contact is not issue rather sensor .
I agree with Ravindra. A driverless car should have a series of sensors that replace the visual contact. In some sense, sensors are the "eyes" of the car; in case of humans, nerve messages from the eye are processed by the brain, while in case of car, messages from sensors should be processed by a special algorithm that would fuse the different messages from the various sensors. See for example "A General Approach to the Fusion of Imprecise Information" by Yager, or an application carried out by myself and Mario Marinelli "Fuzzy Data Fusion for Updating Information in Modeling
Drivers’ Choice Behavior"
The accident rate on roads was fairly low because human drivers do not make maneuver mistakes at the same time and one would give way to the other politely. However, when two computer-driven cars meet, a mistake on one side would cause more mistakes on the other side.
Vehicle-to-vehicle and vehicle-to-infrastructure communication will augment these limitations somewhat. Last week at the Automated Vehicles Symposium Google self-driving car head Chris Urmson showed how vehicles can perceive the world around him. He's also done a TED talk on it.
http://www.ted.com/talks/chris_urmson_how_a_driverless_car_sees_the_road?language=en
- Badges
- Science topic