In Ghostbusters, Egon gives the other ghost hunters an important warning for the operation of their proton packs: don’t cross the streams. Why? Because it would be bad. Even “all life as you know it stopping instantaneously and every molecule in your body exploding at the speed of light” bad. Total protonic reversal. Definitely bad.
Later in the movie, it turned out to be okay to cross the streams when they faced the movie’s biggest bad guys, and it saved the day. The gate to whatever other dimension Gozer the Gozerian came from was forced closed by all the protons, and the Destructor, who had appeared in the form of a large marshmallow man, melted from the flames and heat. Awesome.
But it turns out that automated cars might face a similar problem: crossing the streams is bad. And by streams, EE Times is talking about cars’ radar sensors interfering with each other, and this can cause minor errors at best and total radar blindness at worst. I think we can all agree that having a car rely on radar, and then having that car go blind on you while you’re going down the highway, would be a bad thing. You probably won’t face total protonic reversal, but you could end up dead just the same.
Radar works by sending out radio waves and then listening for the echo of the radio waves that bounce back off of things. It’s not a perfect system, but it’s useful at seeing where things are and how far away they are. It’s not the only way for a car to know this, as vision-only systems like what Tesla is working on and lidar (light radar) systems most other AV developers rely on provide great alternatives.
The problem is kind of like walking around with a bunch of other people in the dark, and everybody has a flashlight. Except instead of carefully and methodically pointing one flashlight around to where you’re looking, the people in the crowd have a bunch of bright flashlights pointing every which way so everything around them is lit up. This approach to using flashlights might work okay for one person, but the problem is that your flashlights are pointing in other people’s eyes, and making their flashlights worthless because you’re blinding them. Also, their excessive flashlight setup blinds you, too, so everybody is now just as blind as they were with no flashlights at all.
Martin Duncan, general manager of ADAS and ASIC division of ST Microelectronics, told EE Times, “The fact that we have now 25% of new vehicles with radar systems, it is already an issue. If you try to capture in real time road conditions, it is very easy to see transmissions from multiple vehicles. As we are all using the same frequency band this, will potentially worsen as fitment rate increases.”
While the suppliers of automotive radar systems have been concerned about this for years, it’s largely been their problem to work on filtering out all of the interference from other radars. The suppliers, along with government agencies in Europe and the United States have been studying and worrying about the interference issue for years, but the automakers and those developing overall autonomy and driver assistance features haven’t had to deal with the issue so much themselves.
Instead of working on coordinating different radar systems so they don’t blind each other (like when I tell my kids to keep the flashlight pointed away from my face at night), the current industry standard for radar is to just be random and hope it works out. The radars use random frequencies and timing so they don’t hurt each other as badly, but this approach won’t work in the future when most cars on the road are using it.
NXP Semiconductors paints a grim picture, but also has a solution:
“… Systems that operate well in environments with few other radars may suffer significant degradation of performance in radar congested environments. The results of the study show, levels of interference based on operation of current systems in congested environments will be significant. In scenarios with many vehicles operating radars in the 76–81 GHz band, the power from other radars will likely exceed the power of echoes from targets needed for specified performance, by several orders of magnitude.
“… Eventually, to support a high market penetration, some form of agreement between the manufacturers will be needed to more effectively share the sensing resources in a fair way. This last step means that all the players in the market will have to sit together to define a standardized way to access the channel while at the same time maintaining the possibility to have differentiating sensing performance.”
The problem is that most in the industry don’t want to go through all this hassle. They find it easier to keep working on improving their own radar to filter out the interference to avoid being blinded. This approach works today, and will work for a while, but could end up being overwhelmed by congestion on the frequencies if enough radar-equipped cars get on the road.
EE Times says the NHTSA thinks there are two big problems with radar that will be very difficult to solve:
In the case of traffic on a two-lane highway, assuming that the radars use randomly selected carrier frequencies, NHTSA predicted that “an automotive radar would encounter power from other radars far greater than the echoes of its own transmissions needed to track other vehicles. The interference approaches four orders of magnitude, or nearly 40 dB, greater than echoes typical of a reference target, as specified for the system.”
In radars that face rearward (as in blind-spot detection systems), “these units are vulnerable to the direct arrival of forward collision avoidance radars that utilize higher power and antenna gain.” The study said, “Our analysis shows these units could experience interfering power from a forward collision avoidance radar that is nearly five orders of magnitude, or 50 dB, greater than the reflections from their specified reference target.”
If the light gets bright enough, even closing your eyes at the right time won’t help. The signals could just be too strong.
Even lidar isn’t immune to problems like this, but the brief pulses they send make meaningful interference much less likely. Ultimately, vision is going to prove important on autonomous vehicles, even if only used as a backup to lidar and radar sensors, because they will prove to not be immune to interference.
Featured image by Tesla.