Throughout the last few years, we have witnessed the race to market for seamless autonomous driving technology within the shortest possible timeframe. It’s clear that one of the biggest hurdles lies in preparing autonomous vehicles (AVs) to navigate the many uncontrollable and unexpected obstacles that face drivers on any given journey.
Real-world testing is of course the ideal scenario. The reality of this happening successfully, however, falls short of producing enough quality, representative, diverse and detailed data to properly train the AI components of self-driving vehicles. As I’ve discussed in previous posts, the very nature of the real world makes it impossible to determine every eventuality. The driving experience is fraught with various unforeseen obstacles, unexpected traffic delays and detours, and sudden changes in weather conditions. In order to prepare autonomous vehicles to properly navigate and react, it is imperative the industry looks beyond a real world-only testing programme.
Playing With Different Scenarios
So if real-world testing falls short, what is the alternative? The industry is quickly discovering that a game engine-based, virtual reality environment offers the best solution. Why? Because it is cheaper, faster and safer than experimenting with actual vehicles. Also, scenarios can be continuously repeated – something that would either hardly ever happen in real life or would pose an intolerable risk to human drivers.
How could you, for example, test things such as hitchhikers in the emergency lane, unexpected roadwork, an impaired driver, or even a sudden traffic accident or weather change? These occurrences that most any driver has experienced at one time or another are, by their very nature, happenstance and impossible to predict or test.
Applying a digital setting for AV testing brings the ultimate benefit of engaging AI for scenarios that simply cannot be replicated under real-life conditions.
Even if you set aside the issues around probability and repeatability, covering thousands of miles would take a massive fleet and a significant investment in time. Alternatively, a set of high-performance computers gets the job done within an hour. Simulating cameras and sensors, in real time if necessary, can cut testing time tremendously and save big dollars.
Prepare AI To Fine-Tune Motion
This highlights another key issue — motion planning. According to current consensus in the industry, machine recognition software can be trained by feeding AI images or footage captured in traffic. The same doesn’t apply to motion planning, however, since the very nature of a moving car influences variables in your surroundings. This is why simulation-based AI is the only way of training around motion.
To understand how this works in the real world, consider the Arc de Triomphe in Paris, one of the craziest roundabouts in Europe. Traffic from 12 major avenues feeds into it, and drivers navigate in six lanes without road markings. French drivers have a choice of auto insurance policies – those that exclude and those that include navigating the Arc de Triomphe.
The very nature of this treacherous roundabout requires drivers to inch rather instinctively, and it would make little sense risking accidents just to train AI on the spot. By contrast, a simulated algorithm can prepare AI to fine-tune motion planning specifically to navigate the Arc de Triomphe roundabout or any number of hotspots all around the world.
Real-World Testing Isn’t Always On The Bench
It’s important to remember that a gaming environment is not the cure-all solution. Real-world testing, despite its shortcomings, still plays a critical role in autonomous development as simulation often lacks the kind of variability that can be found only in real life.
For the self-driving industry, however, it is perfectly fine if around 90-95% of testing takes place in a simulated environment. Observing that ratio is crucial to reaching full autonomy in a timely manner. Smart developers are discovering that game-like simulation provides the smartest, safest and fastest way to put self-driving vehicles on the road.