Featured LIDAR business practices

Not everything that should be avoided, has a contrasting temperature.

 

Remember that GPS is not good enough to be a sole means of navigation.

I've been watching GPS on board commercial flights for a while, when not specifically prohibited, it is a great way to identify and get oriented to landmarks and topography. In 2025, I found it jammed or spoofed on four separate flights. Within sight of Kalingrad, the GPS said the plane was moving backwards at about automotive highway speed. News reports indicated many on-going GPS anomalies in the Baltics region, impacting both aviation and marine vessels. Over Myanmar, both directions with a multi-GNSS receiver, it placed us way off track, and seemingly looping many times around a medium-sized traffic circle at 26 mph, and very curiously, each 'satellite signal' was exactly the same signal strength, which never happens on home territory.

Various websites list or map numerous locations where GPS is known to get monkeywrenched. Certain state actors can do it nearly anywhere at any time.

The radar dynamic cruise in my '24 Toyota has displayed no problems with expansion joints or road work, just a very few anomalies from vehicles in adjacent lanes in curves.

I haven't seen anyone suggesting radar to be good enough to be a sole ADAS sensor type.

 

Seeing in the dark: Unlocking AI in next-gen radar applications | DC51121 | GTC DC 2025 | NVIDIA on-demand

 

Curves can lead to issues like that with the Subaru, and that's cameras only in for the front. Also had the curve issue with snow banks recently.

Radar only might the case when the ADAS comprised of just adaptive cruise control.

 

We weren't flying anywhere near Ukraine or any international war. The first incident was in peaceful airspace over open Baltic Sea waters, between Vilnius and Copenhagen. Searching for news later, there have been tens of thousands of GNSS anomaly reports from across the region over the past several years, including reaching over Sweden.

Myanmar does have an ongoing civil war in some regions, e.g. the north, but civil air traffic is still overflying other regions, we were over the south. There was no active war in the western Pacific waters where another anomaly was seen.

The first noticed incident was on my older hiking unit, which receives both GPS and GLONASS. The multi-GNSS system carried later was simultaneously tracking GPS, GLONASS, Galileo, BeiDou, and QZSS while we were traveling in the Asian region.

Inertial navigation may be good enough for brief flights of higher-yield munitions, but I haven't seen any indication of it being good enough for landing aircraft, especially after extended flights. Aviation use requires augmentation from Instrument Landing Systems and other means.

= = = = = =
I've found that the electrochromic window shades in B787s and A350s seriously hamper GPS/GNSS use. They also give cabin crew the ability to deprive window-seat passengers their views of the outside world.

 

This screenshot is from this presentation I linked. It is amazing how radar can see the invisible. However, it requires extensive digital signal processing (DSP), and Nvidia is working to integrate that with AI.

 

No one here is holding LIDAR as some 'great white hope'. We just acknowledge that the technology has been steadily improving, and don't hold past performance up as reason to completely discount it for Level 5 ADAS.

Despite its impressiveness, Tesla still hasn't taken the step to certify FSD as ADAS 3.

Tesla no longer uses sonic sensors.

 

You seem to promote Tesla FSD even more than Elon Musk and his engineers. Even Elon Musk said Nvidia and Mercedes–Benz were too ambitious to aim for Level 4 because they themselves thought it was too difficult to attain, but he encouraged them for trying.

I see no point in your arguing every day that FSD is great and there can’t be anything better. Haven‘t you heard something called AI and the companies working on it?

 

Elon Musk: We will inhabit Mars in no time. Elon Musk later: We will inhabit the moon sometime in the future.

Elon Musk: Tesla can self-drive. Elon Musk later: Tesla can self-drive with driver supervision. We tried but couldn’t achieve Level 4, and we think it is nearly impossible.

 

OK, well, at least you are admitting that FSD has problems.

 

I really doubt lidar is required for level 5 ADAS. I also don't trust waymo is there yet either - it and tesla robo taxi are ADAS level 4 right now. There are a lot of waymos in my area and they screw up traffic and were passing school busses with stop signs on until the city stopped them from operating during those times. I notice the waymos more because they look different, sound different and there are a lot of them, many more than I think there should be until they get the software better. It is much easier to do self driving in a small mapped area, that is how teslas can be level 4 like waymos, they are not encountering new places. waymo does worse as i see it on construction. Waymo has now admitted to using remote workers in the philopenas to help the make decisions for the waymos.

Supervised FSD beta is a little different beast than the robotaxi software. No one has level 3 driving on all of american roads. Mercedes has it on limited roads in 2 states, but their new system will be level 2+ with lidar.

Lidar can make things smoother and safer. The problem right now is cost and compute, but that is being worked out. Vision is the most important sensing but there is more information to add with radar and lidar. Is an object small or is it far away. radar can do that quickly and see through bad weather. lidar can see more details but not as far and not as well in bad weather. All 3 might make a better less compute intensive neural net. It does need training data. Tesla has the most when it comes to vision but only a little lidar data.

 

A review I read on the OpenPilot system stated it has a tutorial that the driver needs to go through before they can use the system. The car manufacturers are just leaving it up to the owner to read the manual or learn through trial and error.

I'm not making that claim. Just the point that you can't discount it based on past performance. If we did that, we wouldn't be using Li-ion in our EVs now.

Musk and Tesla doesn't play fast and loose in promoting its capabilities, and then hides behind 'it's only Level 2' when someone crashes while using it? Personally, I don't care how well it actually works. My issue is with the culture they built around it that can lead to irresponsible use of it.

Lidar can make things smoother and safer. The problem right now is cost and compute, but that is being worked out. Vision is the most important sensing but there is more information to add with radar and lidar. Is an object small or is it far away. radar can do that quickly and see through bad weather. lidar can see more details but not as far and not as well in bad weather. All 3 might make a better less compute intensive neural net. It does need training data. Tesla has the most when it comes to vision but only a little lidar data.[/QUOTE]As I aid it is improving. They all are. The OP implies Lidar isn't. Going back to Waymo, they use all three sensors. We, as outside observers, can't pinpoint which sensor is at fault for issues, or if it is in the software.