Self-driving cars are going to change many things about our daily lives, and simulation technologies are right at the center of that innovation.
When you think “self-driving car,” you probably envision a car mounted with tons of sensors and computers that drives itself down the road. While that is certainly the external picture of the industry, self-driving cars are essentially real-time simulation engines. To some extent, the self-driving cars of today already implement simulation on a real time basis. They simulate accidents, simulate other drivers using predictive algorithms – self-driving cars use models to predict the world around them.
Shifting from cars running simulations in real-time to avoid current events, let’s look at some simulation technologies currently in use that can be used to generate the code for autonomous cars.
Ever wonder how self-driving cars will make future moral decisions? For example, if a car can predict and knows that it will either hit a mother and a child or hit a father and a son, how does it make that decision? Programming a machine to weigh human life is not easy, and frankly, it’s not something I want my car to do organically through Artificial intelligence. Research is actually being conducted by MIT to help program morality algorithms for self-driving cars, and you can participate! Engineers have devised what they call the “Moral Machine” to get feedback from human users on which not-so-good outcome would ultimately be better. If you’re interested in playing the game (I highly recommend it) you can check it out here – after you’re done reading this awesome article of course. It is this research that will be programmed into the simulation algorithms that autonomous vehicles will use to avoid accidents or optimize crashes in the future.
One of the biggest arguments against self-driving cars is that they are just going to cause more accidents and there’s no way they could drive like I can with all of “my” experience. Natively, we have been programmed to think that our sense of intuition and feeling can’t be programmed into a machine, but you’d be wrong. Simulation and AI are major proponents as to why autonomous vehicles can easily be a better driver than.
Self-driving cars don’t have to be perfect, they just have to be better than humans. Considering humans are actually terrible drivers, this isn’t actually that hard. By letting an AI algorithm run endless simulations of traffic patterns, it can get a pretty good sense of “intuition” as we might call it. Not to mention that when autonomous vehicles are implemented on a wide scale, your car will know what every car is doing in, say, a 1-mile radius around you.
AI autonomous programs don’t even have to watch real traffic conditions, they can learn how to drive by playing Grand Theft Auto. That is not a joke, it is actually happening. Many developers of autonomous technologies are setting their programs loose in GTA to test their ability to self-drive without actually risking any real damage, according to MIT Review. Believe it or not, the traffic and physics engine in GTA V is real enough for autonomous research. I don’t know if that makes me feel better or worse about hopping into an autonomous car…
Simulation isn’t only limited to programming the self-driving cars, it could actually soon tell manufacturers how many autonomous cars they need to produce. It may even demonstrate how many autonomous cars an urban center needs to have in order to make sure the entire population can consistently get where they need to go without actually owning a car. Bolstering this future of simulation determining car needs, research from the Organization for Cooperation and Development found that 90% of urban cars will be unnecessary when autonomous cars hit the market in large quantities. This number was, of course, found by running a simulation of driving activities based on data from urban centers and finding the redundancies in the system.
Simulation will be involved in nearly every aspect of autonomous vehicles. From CFD analysis on car designs to real-time traffic simulations allow the AI systems to make choices and develop driving habits. Simulation is at the core of it all.