Autonomous driving — Challenges

Communication technologies have conquered a large part of human daily activities; from the way we work, share information, keep the agenda, study or enjoy our hobbies. Some of those tasks are especially complex and are only performed by humans, e.g.: driving a vehicle… Or I should say, were only performed by humans. We have all heard about the Google car or the Tesla autopilot, just to name the popular ones; in case you haven’t, watch these videos:

Although the Google car is not available for purchase and Tesla says that even in autopilot the driver must be 100% alert, this is a reality; this will happen either you’re prepared or not. Big car companies have estimated that for the year 2020 they will be commercializing completely autonomous vehicles, take a look at these links:

Google, Apple and Tesla race to driverless cars

30 Corporations Working On Autonomous Vehicles

10 million self-driving cars will be on the road by 2020

Before we get there, today’s on-board technology in vehicles uses sensors to detect its environment and recognize situations of risk; some of such systems can, autonomously or semi-autonomously: pre-detect collisions, warn the driver, apply the brakes and even execute course correction maneuvers on behalf of the driver. However, there is a number of complex scenarios where human capacities are not fast enough to be aware of the threat and react promptly; or others, where a single vehicle applying the brakes is not enough to avoid a collision.

Let us imagine a situation where two vehicles are in collision course, in a two way street; both of them could be capable of detecting the danger and generating an action plan to avoid it. However, if these cars only use data provided by on-board sensors, the plans they will generate individually might involve the execution of evasive maneuvers that are in conflict with the other car’s plan; therefore, they could end up creating a new collision situation, derived from the previous one. Even if both cars are able to detect the new danger, they will have less time to react to it. So, information of what the car can see with its sensors, and what it can deduce the other car will do, is not enough to produce an effective collision avoidance maneuver. Moreover, what if sensors in one of the vehicles don’t recognize the danger, like it tragically happened some weeks ago:

Tesla autopilot driver ‘was speeding’ moments before death

The use of communication technologies to build an environment in which vehicles cooperate to avoid a collision, is the response to this kind of situations; even more, such cooperation is required for scenarios in which vehicles have not line of sight with each other. Therefore, more intelligent vehicles can take advantage of this capacity to inform others about its presence, and even to send information associated to its position, speed, direction, etc. In the near future, applications will use Vehicle to Vehicle (V2V) communication to share data, coming from intra-vehicular sensors, with vehicles in the vicinity, to calculate the imminence of a collision, and to create cooperative plans in which several vehicles act to avoid it.

So, we have 3 and a half years to produce vehicles that are not just able to drive themselves, but are also able to talk to others and coordinate how will they pass over an intersection or how will they prevent a collision from happening. Our cars will be able to change planned routes if up-to-date information, shared by others, indicates traffic jams or bad weather conditions. In the meantime, we must stick our hands to the steering wheel and dream about the day in which we read a book to our car while it drives us home… but, please, dream only when the red light is on.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.