Robo-Cars and the Future of Driving
At first, the concept of cars as robots may seem a little far-fetched. However, you’ve probably heard of – or maybe even seen – one of these robotic cars at work in the form of self-driving cars. Believe it or not, self-driving cars are actually considered robots, and are even called robo-cars in some industries and countries. In the US, many companies are hard at work on transforming these robotic cars into a viable mode of transportation.
So what, exactly, is a self-driving or robotic car? According to one widely-accepted definition, there are actually six different levels of automated, or driverless function. At the very basic level, the car has the ability to warn the user of potential obstacles (lane departures, other nearby vehicles, etc) and even take some limited action to avoid accidents. These features are widely available in most modern retail brands, and can vary a lot from model to model. The next level that is currently widely available is “hands on”, which means the car helps the driver but ultimately shares control of the vehicle. This can look something like cruise control, or even emergency braking if the car detects a danger to the user.
The next tiers of automation (all the way through the sixth level) are far more experimental, and aren’t widely available in most car models. For those seeking to make entirely robotic cars viable, the goal is to reach the hallowed “level 5”, or “steering wheel optional”. A car that can function at level 5 should be able to take the user wherever they want, in any conditions, without the user having to interact with the vehicle at all.
As you can imagine, current self-driving car technology is far from achieving this lofty goal. So how exactly does this technology work in its current form? Well, each company makes their robo-cars slightly differently, but as of now they all operate on the same general principle. First, the car generates a “map” of its position, using laser rangefinding sensors, GPS, and much more. Different cars use different technologies to perceive their environment; some use cameras, while others use radar (radio-wave based vision) and LiDAR (light based vision). There are drawbacks and advantages to each of these; for example, radar is lower resolution but can see in bad weather, while cameras capture a great image but we currently lack the AI to interpret it. Radar doesn’t capture smaller objects as it uses longer wavelengths. LiDAR uses shorter wavelengths and can detect smaller objects and can build an exact 3D monochromatic image of an object. However, they don’t work well in the night time/cloudy weather. Some cars also use microphones to pick up surrounding noise, or signals from other properly equipped cars.
The car then identifies any obstacles on the map, like pedestrians or other vehicles, and seeks to avoid them. After the map is generated, the car is able to generate hundreds of path possibilities, and choose the best one to continue towards its destination. Car computers use different machine-learning based algorithms to make decisions, as well as identifying objects and determining their distance from the car.
In the present day, however, there are many constraints that prevent self-driving cars from taking over; weather, sensor limitations, local and federal laws, insurance, liability, and more. Additionally, AI models are not advanced enough to deal with subpar environmental conditions, or the mistakes and errors of non-automated drivers. Above all, driving is an incredibly complex task that requires a lot of human judgement and observation – much of which robo-cars simply can’t replicate.
If engineers are able to overcome these challenges, there is also a lot of upside to be had. For one, a fully automated traffic grid would be able to increase efficiency, reduce carbon footprint, and improve life for pedestrians and residents. Hopefully, new innovations and breakthroughs will soon make level 5 self-driving cars a reality.
Self-driving cars
Picture Source: continental.com
