Full autonomy is closer than ever, using artificial intelligence will allow vehicles to react to the real-world by taking data inputs of different sensors. This seems like something promising, but it will take some time. The autonomous driving capabilities encounter different challenges, ranging from moral considerations to technical issues (Brandom, 2018).
However, the Society of Automotive Engineers (SAE) defined many benefits of autonomous driving capabilities, especially with the regard to life-long mobility and a decrease of accidents. It is therefore important to integrate artificial intelligence capabilities and focus on further development of automation (Reser, 2018).
As stated above, sensors will enable vehicles to react to its environment by developing a representation of the real world. Developers need to focus on the identification and control of the input parameters send by the sensors. A sensor consists of various radars, ultrasound and cameras. It is however a challenge to discover correlations between AI scenario’s, deep learning, the physical signals and the real world impact a decision will have in real traffic situations (Reser, 2018).
Other researches argue that fully autonomous cars are further in the future than most people realize. For deep learning to work properly, it will require a lot of training data to integrate almost every possible scenario. For algorithms to recognize different scenarios, engineers can use their creativity for the source of data and its structure. However, there is a limit when it comes to how far a certain algorithm can go. This recognition process, also called ‘generalization’, requires certain skills. Research has shown that deep learning is not very good in doing this generalization process, leaving the autonomy companies with many concerns (Brandom, 2018).
- Brandom. 2018. Self-driving cars are headed towards an AI roadbloack. https://www.theverge.com/2018/7/3/17530232/self-driving-ai-winter-full-autonomy-waymo-tesla-uber
- Reser. 2018. How AI will help pave the way to autonomous driving. https://www.electronicdesign.com/test-measurement/how-ai-will-help-pave-way-autonomous-driving
Hi Ghofrane,
Great read!
In your last paragraph you argue that autonomous cars are further in the future because a lot of training data is needed the integrate every possible solution. But there are different methods of AI that work better for autonomous driving, allowing the car to avoid these limitations right? I believe it isn’t such a big concern as you state it is.
The bigger concern is whether the law will allow it. A large amount of a governments budget is collected through ticket (such as speeding or riding through a red light). Would a government really be willing to just say yes to perfect driving, knowing it would miss out on a lot of “income”. Once these cars are allowed, how is the government going to make up for it? Higher taxes on cars?
Dear Ghofrane,
Thank you for your interesting post! I fully agree with your article. I think we are ahead of time in using self-driving cars. First, we need to optimize automatic parking since there are still a lot of troubles with that. There are a lot of youtube videos online about cars not able to park automatically in a proper way. This means that this technology is far from perfect. Self-driving cars could be really dangerous if they are not working properly. This could involve bumping into people while driving as well. If in the future this technology is fully optimized, I would say self-driving cars are the future. But until then a lot of research and experimenting needs to be done.
Kind regards,
Juju Rovers
Dear Juju,
Thank you for this comment! I agree with you when it comes to the optimalization of automatic parking, as it is still not perfect. Although autonomous vehicles are closer than ever, it’s still needs a lot of research to ensure that the vehicle would be totally safe to use. I hope this would happen in the near future as it would change the way we travel!
Kind Regards,
Ghofrane Bouaissaoui
Very interesting post! I agree with you that there is a lot testing and development needed until there will be fully autonomous cars. However, Waymo (Alphabet´s autonomous car project) is currently testing its algorithms on more than 25,000 miles simulated driving per day. In total they tested them on more than 2.7 billion miles alone in 2017. Furthermore, I would like to stress that there are more than one million deaths per year as a result of road traffic crashes and dozens of million people get injured. So why should we only accept autonomous vehicles that are 100% accident free?