The use of wireless technologies is an integral part of modern life. Radio, TV, Wi-Fi and Cellular technologies are part of the backbone of the economy. However, even with these established technologies, there are issues with providing adequate coverage. Interference from both man-made and natural sources is a concern. Today, when issues occur, the result is largely inconvenience. Autonomous Vehicles (AVs) use a whole host of wireless technologies as a basic part of perception and communication. If not carefully handled, there is the potential for significant safety issues because misbehaviour of AVs can cause significant physical harm.
First, is this really a problem? An incident with an Intel Mobileye AV demonstration in Israel provides some insight. In this incident, Mobileye’s car ran a red light, and according to Mobileye, this was due to ElectroMagnetic Interference (EMI) between a wireless camera used by the TV crew and the traffic light’s wireless transponder. This happened at a well-orchestrated demonstration. One would expect the real-life operation to be the worst. In general, EMI is a real problem which is easy to ignore because it is invisible. The reality of EMI’s intersection with AV operation opens up a whole host of questions.
The first question is: How big of a problem is this? The answer seems to be that we do not quite know. There are three forms to consider:
- EMI with Infrastructure: The FCC has regulated EM activity for the last 85 years. There are lots of potential sources of man-made EM noise (airports, radio stations, power plants, etc). The critical question is: Do the current regulations around these entities account for an EM-sensitive AV driving nearby? Seems doubtful. What exactly does the EM noise map look like on American roads? It would be nice to know. Beyond the FCC, the vast majority of transportation civil infrastructure regulators have never cared about being conducive to EM issues in the past. Do certain shapes (e.g., tunnels, overpasses) have to be handled with more care? Do we need to avoid certain reflective surfaces? Seems likely.
- EMI between AVs: Each AV uses active transmission of EM energy with systems such as radar and lidar. When we are in an urban environment with a high density of AV cars, how do we know we are sensing our AV’s responses? How do we know that congestion/interference/reflection will not be an issue?
- EMI with Weather: As we all know, weather phenomenon can have an impact on communication systems, and certainly impacts sensor systems. Weather is not controllable, so how will weather-related interference be handled. After all, storms do not care about the FCC reservations of frequency bands.
The second question is: Can we build the infrastructure to support it? There are those who believe fast Vehicle to Vehicle/Infrastructure(V2X) technology combined with cloud intelligence can solve many issues. In this world, beyond getting “loose” information on the location (GPS), there are those who believe that AVs can be managed in a “live” manner to a point where traffic lights will become unnecessary. Given the difficulty of providing existing technologies such as cellular with full reliability, is it possible to provide the level of reliability required for AVs? If so, what will it cost? Is it worth it? The recent experience with DSRC does not bode well for this direction. In a recent decision, the FCC reassigned a large part of the DSRC band after a lack of progress over 20 years. As discussed in a previous article, the underlying reasons may well be connected to the unique structure of the transportation economy, but even at a pure technical level, the challenges are daunting.
Overall, the use of wireless technologies is very attractive for AVs but also brings along a whole host of potential issues. Moving all the pieces to make it all work will be a significant challenge. Perhaps, Elon Musk is right. Perhaps AVs should be limited to camera systems? After all, this aligns most closely with the human-based infrastructure which already exists and is actively maintained.
All credits for thos article to Rahul Razdan in original link below: