The news of the technology can be very eye-catching. For example, if we talk about autonomous vehicles, soon in our imagination as we are transported to a future where transit is a thing of the past, where we can be relaxed reading, or just listening to music, while the auto by itself deals with the environment of a sea of cars. Soon our imagination takes us to an ideal world where the peseros are a thing of the past and it doesn’t get the brava, do not commit a thousand outrages (and violations) that are never punished.
But going back to the real world, perhaps the automobile self-employed could prevent someone from last of the glasses up to your car and tramples someone, or well, that simply the driver being distracted in a given moment. In the united States die about 40 thousand people per year in traffic accidents and 90% of these accidents are due to human error, according to the National Highway Traffic Safety Administration of the united States.
It is clear that the autonomous vehicles not given sufficient confidence today. For example, a test car of Uber struck a woman on the street killing her, which is recorded as the first fatal accident of a vehicle driven only. So, simply this fact generates sufficient distrust of the broad public.
As is, apparently, the technology does not think to stop by these drawbacks, and the cars self-employed will come to us when the future reaches us. But the reality is that may never come to be among us. Nidhi Kalra, an robotista who co-directs the Center for the Rand Corporation for the decisions that are made in events with uncertainty, indicates: “The technology is constantly updating. So, we have this new self that is handled alone, we have this product, but with the software updates there is a new car practically every week”.
The issue is therefore a problem of software for autonomous vehicles. More than half a million lines of code to do that will eventually have cars that can navigate without human help. This includes tracking systems, maps high-definition so that the car knows where it is precisely, as well as perception systems that help determine what there is around the car. What are people that see the car? How can any of them suddenly want to cross the street at an inopportune moment? And if all this were not enough, there are planning systems that synthesize all the information to be able to go from point a To point B without any problems. Finally you need electrical and mechanical systems to which the accelerator is pressed and the car is moving. Think, reader, reader binaries: driving a car is a problem extremely complex, and not only for the technical part, but all over the environment, ethics and what it entails.
And if all this weren’t enough, think of the conditions of the terrain or the weather or worse still, in the idiosyncrasies of the drivers, that changes from city to city. Or consider the difficulties of driving at night, or when there is fog. The variables are multiplied and the complexity increased. So you can imagine that if the upgrade of a phone is very often, the software of the autonomous vehicle should be done even more times by the complexity that it is attacking.
“Any product can be improved with time,” says Mike Wagner, co-founder and CEO of Edge Case Research, which helps companies robotic to build software more robust. “This is the cycle d life maintenance of any system,” he says. So, possibly if a car is handled only required to make a stop at the nearest café, then will probably require that you update your software. Or with the time probably the autonomous vehicles require you to learn new patterns of traffic, or the weather changes that are not necessarily going to always keep equally. In any of these cases will possibly need to update the software to reflect all these new issues.
“The environment is not static,” says Forrest Iandola, the CEO of DeepScale, that builds systems of perception. “Even if you, in theory, has a perfect system for a certain locality, this becomes obsolete,” he says. And under this idea it is clear that the autonomous vehicles can face unexpected situations by anyone. For example, what if you escaped to an undetermined number of tigers in the zoo? how to anticipate what to do? You will then need time to train the system in this new environment that was unimaginable and then, update the software.
But perhaps the worst problem is with the security environment necessary for autonomous vehicles to be sufficiently safe, both for human beings to address as those who are on the streets where these vehicles are in transit. Suppose that we have a program to stop automatically or, a system of automatic handling adaptable. This obviously is software and this should be tested. Clearly these programs examples can be demonstrated as a functional step-by-step, but when we speak of a complex system really, as the autonomous vehicle, the scale of the system tends to grow and update the software and in addition, prove that it is functional, it can be much more difficult.
Since then all these problems can be attacked, but they require a lot of work. For example, if an autonomous vehicle has a system of geolocation and use in certain conditions, you must ensure that nothing is failing because this can cause difficulties. And to think that I may fail is the best scenario because the systems do not have word of honor. So then, we need to implement systems of self-diagnosis to make sure that all of the systems to do the task at least as it was originally proposed.
All scenarios does not mean that they finished the project of the self that can be handled alone. However, it is clear that the development will be much slower than expected. Probably let’s first look at autonomous vehicles that travel short distances in certain areas nothing more, but with the time, with the experience that these first situations, we can move forward. What must be clear is that the autonomous vehicle is still far from the real possibilities.