First Jerry rig everything, now mark rober 🧐

[CeloTru]

Great reply! First off I agree with you I am a dufuss. However you must have not watched the follow up video with Mark Rober where he said they went back and cut the foam out months later and reshot the whole thing so it gave a more dramatic effect…

Sponsored

 

[dalton108]

What kind of dufuses are on this forum? It WAS pre-cut because it was solid and any 1st year engineering student that understands even the most basic concepts around vision-only based ADAS system would have know there was a good chance it would hit the wall. Regarding the pre cut wall… Do you really think the cartoon like cut out was meant to fool people into thinking that styrofoam would fail that way when it was hit? It had to be pre-scored in order to reduce chance of damage to the vehicle if the most likely of scehappened. So why not make it a fun pattern? At 40 mpg the styrofoam could have caused major damage without a way to reduce the force of impact by providing mechanical “slide” to absorb the energy. You know… because of that thing called physics? Remember Columbia?

The science behind it…
Hitting a 4-foot-thick Styrofoam wall at 40 mph could cause damage to a vehicle, depending on the structure of the foam and the design of the car. Styrofoam absorbs impact by compressing, but it is not as effective as materials specifically engineered for energy absorption, like SAFER barriers or crumple zones[1][7].

If the impact area of the Styrofoam wall is pre-cut to allow the hit section to slide out, it would reduce the force of impact by increasing the collision duration and decreasing peak impact force. This principle is similar to how crumple zones or deformable barriers reduce damage in collisions[6][9].

Sources
[1] Why is bumper styrofoam still in tact after rear end accident at 30mph? https://community.cartalk.com/t/why...n-tact-after-rear-end-accident-at-30mph/81005
[2] [PDF] The Use of Impact Force as a Scale Parameter for https://ntrs.nasa.gov/api/citations/19950022058/downloads/19950022058.pdf
[3] What is the impact speed of 2 identical cars going at 40mph ... - Reddit
[4] Light impacts still feel harsh, is the foam in the front & rear bumpers ...
[5] Impact force reduction by consecutive water entry of spheres https://www.cambridge.org/core/jour...y-of-spheres/1F76DB78B53CCB244257451F22A244ED
[6] Car Crash Calculator https://www.omnicalculator.com/physics/car-crash-force
[7] The SAFER Barrier | University of Nebraska–Lincoln https://mwrsf.unl.edu/saferBarrier.php
[8] [PDF] Lab #4. Reducing peak force in an impact by crush- ing foam https://astro.pas.rochester.edu/~aquillen/phy141/labs/crush_lab.pdf
[9] Calculate the impact force in a collision - GIGACalculator.com https://www.gigacalculator.com/calculators/impact-force-calculator.php

For those who believe that pre-cutting the Styrofoam is a setup rather than a safety precaution, it’s important to explain the reasoning behind this design choice:

1. **Energy Absorption and Impact Reduction**: Pre-cutting Styrofoam allows the material to deform or slide out upon impact, which dissipates energy more effectively and reduces the force transferred to the vehicle. This is a standard engineering principle used in crash barriers and safety systems to minimize damage and injury[2][4][7].

2. **Safety Design, Not Deception**: The pre-cutting is done deliberately to enhance safety by controlling how the Styrofoam behaves during a collision. Without this design, the foam might resist more rigidly and unpredictably, increasing the risk of damage[3][6].

3. **Practical Precedent**: Similar methods are used in packaging fragile items, where foam is custom-cut to absorb shocks and vibrations, ensuring protection during transit. This demonstrates that pre-designed deformation is a practical safety measure, not manipulation[1][4].

This approach prioritizes safety by reducing harm, aligning with established practices in both engineering and packaging industries.

Sources
[1] The Advantages of Custom-Cut Foam for Shipping Your Products https://solutions.sterlinghouston.com/blog/custom-cut-foam-for-shipping
[2] What is Styrofoam? The Ultimate Guide to Styrofoam - Epsole https://epsole.com/what-is-styrofoam/
[3] Understanding XPS Foam: Uses and Advantages - Atlas Stucco https://atlasstucco.com/blog/xps-foam/
[4] 4 Reasons Why Styrofoam Protective Packaging Works https://solutions.sterlinghouston.com/blog/4-reasons-why-eps-foam-works-as-packaging-protection
[5] Five Benefits of Foam Plastic Insulation https://www.plasticsengineering.org/2024/08/five-benefits-of-foam-plastic-insulation-005971/
[6] Every Type of Packaging Foam Explained: Pros and Cons - Creopack https://creopack.com/en/blog/types-packaging-foam/
[7] Top 5 Things You Need to Know About Polystyrene https://www.engineeredfoamproducts.com/news/top-5-things-you-need-to-know-about-polystyrene/
[8] Expanded Polystyrene Ordinance - LA Sanitation - City of Los Angeles https://sanitation.lacity.gov/san/f...s-lsh-wwd-s/s-lsh-wwd-s-sr/s-lsh-wwd-s-sr-eps

What in the poodle-haired hell are you talking about? Follow up: Who cares? This is out and out fraud!

Even the damn still shot that you used shows that the car is coming through a rectangle approximately the same size as the car before any movement in the cartoon cut out! And it very clearly is not a solid wall but rather an amalgamation of several smaller squares!

What are you smoking?!

 

[Deleted member 22486]

In a nutshell- Mark’s test was completely logical and he was out to validate a hypothesis most engineers would have agreed with- a vision only based ADAS cannot cope with a 2D image of a road.

Jerry’s test however was BS and not valid- applying 11,000 pounds of vertical load instead of horizontal load to the hitch. Trailer hitches are rated for towing capacity, which refers to the horizontal force they can handle when pulling a trailer. Applying vertical load tests a different aspect of the hitch's structural integrity and is not representative of its towing capacity. Vertical loads are typically limited to much lower values, such as tongue weight, which is generally 10-15% of the gross trailer weight[2][6][8].

Sources
[1] Method for testing a vehicle trailer hitch - Google Patents https://patents.google.com/patent/US3575046A/en
[2] How to Verify You Are Towing Within Correct Ratings and Guidelines http://www.weigh-safe.com/towing-sa...towing-within-correct-ratings-and-guidelines/
[3] Tow Hitch Test Rigs - CFM Schiller GmbH https://www.cfm-schiller.de/en/test-rig-technology/tow-hitch-test-rigs/
[4] The Ultimate Guide to Safe Heavy-Duty Towing http://www.weigh-safe.com/towing-safety/the-ultimate-guide-to-safe-heavy-duty-towing/
[5] Testing - B&W Trailer Hitches https://www.bwtrailerhitches.com/testing
[6] Cybertruck Owner's Manual Says Its Hitch Is Only Designed ... - Reddit
[7] CURT Engineers, Fabricates & Rigorously Tests Our Hitches in the ... https://www.curtmfg.com/curt-testing-standards
[8] Why only 11000 lbs towing? | Tesla Cybertruck Forum https://www.cybertruckownersclub.com/forum/threads/why-only-11-000-lbs-towing.13474/

 

[dalton108]

In a nutshell- Mark’s test was completely logical and he was out to validate a hypothesis most engineers would have agreed with- a vision only based ADAS cannot cope with a 2D image of a road.

Jerry’s test however was BS and not valid- applying 11,000 pounds of vertical load instead of horizontal load to the hitch. Trailer hitches are rated for towing capacity, which refers to the horizontal force they can handle when pulling a trailer. Applying vertical load tests a different aspect of the hitch's structural integrity and is not representative of its towing capacity. Vertical loads are typically limited to much lower values, such as tongue weight, which is generally 10-15% of the gross trailer weight[2][6][8].

Sources
[1] Method for testing a vehicle trailer hitch - Google Patents https://patents.google.com/patent/US3575046A/en
[2] How to Verify You Are Towing Within Correct Ratings and Guidelines http://www.weigh-safe.com/towing-sa...towing-within-correct-ratings-and-guidelines/
[3] Tow Hitch Test Rigs - CFM Schiller GmbH https://www.cfm-schiller.de/en/test-rig-technology/tow-hitch-test-rigs/
[4] The Ultimate Guide to Safe Heavy-Duty Towing http://www.weigh-safe.com/towing-safety/the-ultimate-guide-to-safe-heavy-duty-towing/
[5] Testing - B&W Trailer Hitches https://www.bwtrailerhitches.com/testing
[6] Cybertruck Owner's Manual Says Its Hitch Is Only Designed ... - Reddit
[7] CURT Engineers, Fabricates & Rigorously Tests Our Hitches in the ... https://www.curtmfg.com/curt-testing-standards
[8] Why only 11000 lbs towing? | Tesla Cybertruck Forum https://www.cybertruckownersclub.com/forum/threads/why-only-11-000-lbs-towing.13474/

??Lunacy!

 

[Deleted member 22486]

You maybe didn’t watch the follow up video

Great reply! First off I agree with you I am a dufuss. However you must have not watched the follow up video with Mark Rober where he said they went back and cut the foam out months later and reshot the whole thing so it gave a more dramatic effect…

I watched the follow up video 3 times before writing my replies… like any good engineer I do my research.

Mark went back to pre cutting the styrofoam in that pattern as a dramatic effect because it’s YouTube. It also served a mechanical purpose.

Maybe someday the FSD can solve the problem with just cameras, maybe Tesla adds in other sensor types to solve the problem. That said Everyone please get over “the Elon is always right” syndrome and accept the possibility that in its current state, Tesla’s ADAS using vision-only has some Achilles' heals. This scenario is one of them. Mark just monetized it AND used science, physics and engineering to validate his production.

 

[Deleted member 22486]

The whole bias of the video is to fool a self driving car. If you’re naming your title “can you fool a self driving car” then you need to utilize FSD. Autopilot is not self driving. It’s a glorified lane keep assist with traffic aware cruise control and forward collision prevention.

What’s funny the luminar website had in a banner that stated tested by a nasa and apple engineer with clips from the video. But now all that has been removed.

seems fishy still

First of all, It’s not a bias, it’s called a hypothesis. Secondly, if you think it’s about Luminar and/or just Lidar vs vision you have missed the point. It’s about showing the fallacy of relying on one sensor type with computer/software technology available TODAY. With the state of the art TODAY we either need a quantum leap in FSD computing tech to work with just visible light cameras OR vehicles need a robust multi-layered perception system what data that can be correlated together in real time to navigate and keep the occupants and vehicles safe (which probably requires a quantum leap in FSD to do that as well, since sensor fusion is where Tesla and everyone else has difficulty).

The Sensors… Cameras provide visual context, LiDAR ensures spatial precision, and radar enhances reliability in poor weather, together collectively improving safety and navigation. It’s that simple and that was the point of the video. Mark just showed off ONE scenario, where LiDAR would have some advantages over cameras and demonstrated that. It’s a special scenario because it’s one that is deployed on millions of vehicles around the world. It’s one that all of us rely on so it’s especially relevant.

Mark is also a YouTuber. He entertains. He monetizes his work. His medium is video so of course there is video that he won’t use. That said Elon monetizes his work too or are some of you missing that point? I bet if Elon shows us ALL his data on FSD, exposing all its flaws as he knows it, Mark would GLADLY post all of his data. But Elon can say “it’s proprietary” and hide behind a corporate wall.

Re-watch the video, try to strip out the entertainment wrapper and see that Mark uses the scientific method (like any good engineer would) by doing the following:
1. Observation: Identify phenomena or problems that require explanation.
2. Question: Formulate a specific, testable question based on the observation.
3. Hypothesis: (not bias, HYPOTHESIS) Develop a testable and falsifiable explanation or prediction.
4. Experimentation: Design and conduct experiments to test the hypothesis while controlling variables.
5. Analysis: Examine data from experiments to determine whether it supports or refutes the hypothesis.
6. Conclusion: Draw conclusions, refine the hypothesis if necessary, and predict future outcomes.

Maybe it’s time for some of you to step back, and look at the core argument and hypothesis that Mark put together as a VALID engineering study (combined with entertainment as a way to make money) and not some “fishy” conspiracy theory.

Regarding Luminar advertising this on their website (which was debatable advertising on their part I would agree), let’s remember that Elon had the president of the United States, pitching his cars on the lawn of the White House.

The President of the United States.
On the lawn of the White House.
Pitching Teslas.

As taxpayers, which one should we be more worried about?

 

[mongo]

Vertical loads are typically limited to much lower values, such as tongue weight, which is generally 10-15% of the gross trailer weight

Only when stationary.
Otherwise, please explain SAE J684 and Directive 94/29/EC for Europe.

 

[BannedByTMC]

Vertical loads are typically limited to much lower values, such as tongue weight, which is generally 10-15% of the gross trailer weight

All that engineering and you don't understand that dynamic vertical loads on a hitch can far exceed tongue weight?

 

[Mini2nut]

Rober may need to get his own lawyer soon. I read that Tesla is seriously considering taking legal action due to the misleading video content.

 

[jimbobb23]

Can vision only self driving be fooled? Of course. Can it make mistakes? Of course.

These aren't even interesting questions. If you intentionally try to fool most systems you can. It's not interesting.

Here is interesting...in normal driving humans are exposed to can vision only self driving match or exceed humans. Telsa is betting yes and I think they are right. Maybe in AI 8 it will not be fooled but the fact you are asking if it can be fooled is a testament to how good it is. People aren't asking can it drive anymore....we all know it can. The "can we trick it" stage is an amazing place to be and means we are not far from unsupervised self driving.

 

[jimbobb23]

All that engineering and you don't understand that dynamic vertical loads on a hitch can far exceed tongue weight?

weird that none of the cars build them that way then....maybe all the car engineers are right?

 

[mongo]

weird that none of the cars build them that way then....maybe all the car engineers are right?

Huh?
The receiver test spec for an 11k trailer rated vehicle is around 6,000 pounds, not 1,100.

 

[cybercricket]

First of all, It’s not a bias, it’s called a hypothesis. Secondly, if you think it’s about Luminar and/or just Lidar vs vision you have missed the point. It’s about showing the fallacy of relying on one sensor type with computer/software technology available TODAY. With the state of the art TODAY we either need a quantum leap in FSD computing tech to work with just visible light cameras OR vehicles need a robust multi-layered perception system what data that can be correlated together in real time to navigate and keep the occupants and vehicles safe (which probably requires a quantum leap in FSD to do that as well, since sensor fusion is where Tesla and everyone else has difficulty).

The Sensors… Cameras provide visual context, LiDAR ensures spatial precision, and radar enhances reliability in poor weather, together collectively improving safety and navigation. It’s that simple and that was the point of the video. Mark just showed off ONE scenario, where LiDAR would have some advantages over cameras and demonstrated that. It’s a special scenario because it’s one that is deployed on millions of vehicles around the world. It’s one that all of us rely on so it’s especially relevant.

If we limit the context to FSD, there really isn't much that is missing from the vision-only approach at this time. Like literally I can't tell you a REALISTIC example of where the cameras fail and LiDAR wouldn't. And even if you could come up with one, by your own admission one would need to combine the data and the incomplete dataset coming from LiDAR would be still insufficient for the FSD. The [current] inability of vision-based FSD to discriminate against 2D images is only a concern in the context of purposeful deception of these systems and not the normal driving scenarios.

All of my previous comments focus strictly on forward collision avoidance where having a different source of ranging data is needed given the poor performance of vision in adverse conditions where humans generally continue to drive. One common scenario I encounter in the PNW is driving through a hilly terrain where every time you pass through a lower section of the road you go through a dense wall of fog until you get onto a hill again. No human would sit and wait for hours for that fog to clear, so the only way to improve on safety there is to ensure AEB actually works reliably for common types of obstacles even in those conditions.

 

[BannedByTMC]

weird that none of the cars build them that way then....maybe all the car engineers are right?

What's "weird" is you don't understand that they do "build them that way".

 

[DJAlan2000]

I think the thing most people don't realize, especially those that are 'against' Tesla's camera systems and say 'LiDar is better' don't realize that Tesla isn't trying to make a car with the MOST forms of 'detection' and try to get a system that is 100% correct all the time because IT WILL NEVER HAPPEN!

You can put in thousands of dollars worth of Lidar, then thousands more in other 'sensors', and even more in 'radar' and much, much more in cameras... But although all that will HELP prevent accidents, it would prevent 100% of them... EVER!!

Unless you can have a vehicle that can have ALL those modes and then a computer system that can actually USE all those modes at the same time, comparing them all to each other before it makes a decision AND some way for the car to 'jump straight up' if some idiot is detected by the 'system' is about to hit you from the side, front or rear... Without it, you will STILL be vulnerable to get into accidents!

You show me ANY car that can avoid ALL accidents and I will crash something into it...

Telsa is just saying that ITS camera system is WAY BETTER than humans 'two cameras' and can make the decisions FASTER as well...

Sponsored