Cruise Under the Hood 2021 | FULL EVENT
HTML-код
- Опубликовано: 24 июл 2024
- Self-driving technology is on our doorstep-but how do we achieve a vision of serving hundreds of cities and millions of customers with a truly driverless rideshare service?
Scalability is the key to unlocking the power of self-driving in our cities. Watch Under the Hood, where Kyle & other Cruise engineering leaders will take you through an in-depth look at how we’ve built systems and tools to scale, and where we’re headed from here.
Want to join our mission? See open roles at getcruise.com/careers
00:00 - Opening remarks with Kyle Vogt
15:40 - Building the Most Advanced AV with Yun Jian & Brandon Basso
54:50 - Accelerating AV Development with Sid Gandhi
1:04:49 - Creating the AV Ecosystem with Davide Bacchet
1:18:00 - Pioneering the Future of Transportation - Origin
1:43:12 - Interlude with Oliver Cameron
1:46:41 - From Self-Driving R&D to Self-Driving Reality with Jackie Shannon
2:00:34 - Closing and Q&A with Kyle Vogt & Mo Elshenawy
Thank You Everybody for All that you are doing for our Planet Earth.... Peace.. Shalom.. Salam.. Namaste .. 🙏🏻 😊 🌈 ✌ 🌷 ☮ ❤🕊
superb learn something new every day thanks for sharing
this is amazing, very impressed.
A good watch at around 1.75x speed. Especially enjoyed the demonstration of different uncertainties and how to handle them.
Liked them, but liked Tesla's method more... Temporal and Location buffers would have made it so the event at 28:00 or so wouldn't have forgotten about the cars on the right. Tesla also merges all cameras together before analyzing them which seems better. Best is how they are going to analyze photon counts instead of the processed pixels.
Thanks Cruise team for the details, sounds like team build an awesome infrastructure and ecosystems for scientists and engineers to focus on delivering features and not worry about wasting time on non productive tasks. Although, I like use of simulation for test data, map, traffic etc but I am not sure if over reliance on simulation would cover all possible scenarios at scale within expected latency.
It would be great to see a video on blocker bugs and challenges involving complex real time decisions impacting latency. Good Luck!
incredible tech and people. Even more incredible is sharing the knowledge with everyone. Kudos @cruise
Thank you Crisp Imaging - Thank you for your focus on the now and your vision for the future, Customers for Life - Jeff Rabbitt AO Architects, Partner
Very cool!
for sure one of the tech companies that excites me the most out there. I understood that vogt really meant business when they chose SF to run theirs sdv
Love these cars
This is sooo complexxxx.
This is insane !
This is quite impressive! Thank you for sharing your tech stack details!! Although from a high level perspective the sensors stack seems expensive and inefficient (in my opinion), the software stack is something else 🔥🔥
Very detailed engineering! Thanks. One question, The siren sound could be a way to stop the cruise AV from moving? i see the car stopped immediately after a siren sound, rather it should've recognized the police car and the siren and couldv'e made the decision, isnt' it?
It was at crossroads so it just stopped, seems like a brilliant decision I think. If it were at 40 mph then I would say just slowing down would be a good decision.
For the lost keys example you might get always with running the images through an object recognition algorithm and automatically alerting the passenger if anything relevant is found.
Did they say, when it will be available for the public in SF?
102:30 is that actually how good your radars are when there are metal man hole covers in the scene?
Wondering how the Chevy Bolt recall is affecting Cruise, given their entire fleet is based on that platform. Official GM advice is to park fifty feet away from other vehicles, and on a top floor or open level. How has that changed Cruise's operating procedures? How quickly will the battery packs be replaced on these vehicles?
GM is the majority shareholder of Cruise. Although I have no evidence, I can still guarantee that their fleet was as the top of the priority list for battery replacement.
Amazing video!
Cruise vs Bicyclists 😅
We have a great pilot opportunity for this project, with a business model .
How does a vehicle handle getting stuck in snow. Does it detect drifts, high wind areas, black ice?
Good video, although I was hoping you would cover "why" the Cruise sensor approach of using radar, LIDAR, and HD mapped areas is needed and the only way forward. To the best of my knowledge this was not covered in this video.
Because it is kinda obvious that there shouldn't be a risk of a vehicle like.... not seeing a train, or a UPS truck coming out of dark, like what Tesla did. LiDAR is the only concept that assures that you see 100% of all objects around you, no matter what color, and that in rain and snow still better as what cameras can detect at all. It is sociopathic to imagine that you should make full autonomous vehicles which are just guessing what they see and not knowing.
Cruise, please answer these questions as to the scalability of this technology for investors:
1. How many wh/mi increase does this system cost the vehicle? Or put another way, if this Chevy Bolt and a stock Chevy Bolt drive from San Fran to LA, what additional charging time is needed due to aerodynamic and computer consumption.
2. How much time and money does it cost to HD map an area, and what is the time and cost estimate to map the entire US for example.
3. How often do you have to maintenance these HD maps and what will this cost on a yearly basis?
4. What happens if the car gets to an area that is not mapped or has changed since it was mapped?
5. Will this technology work with rain, snow, fog, worn lane markings, dirt roads, temporary construction or other traffic changes, and all normal driving conditions?
6. What is the incremental cost per vehicle for this sensor suite and at what point will millions of these be produced per year?
Thank you for your time! We live at a very exciting time in history!!
Great Video
will this be a subscription or what is cost
Interesting video. Tesla could use some of these innovations, especially around uncertainty, and regressions!
Both ways. Tesla might rely on their buffers too much, Cruise needs Temporal and Location buffers. They shouldn't have forgotten about the cars in the demo they had around 28:00.
All use this. It is a standard approach.
I support your mission. The question is, do I just pay to ride in a robotaxi, or can I OWN a robotaxi and earn revenue from it. Not just shares of your company, but shares of the profit generated by your robotaxis as a form of income. Because the next wave of income when robots take all our jobs is ownership of the robots and revenue from their labor.
I can see a path toward that with your competitors. Let me know when I can buy shares of your robotaxis and earn an income from them.
Is this a self driving training video?
I love you very much
And when lidar dependent, is your system crippled by dense rain and snow? as well as modest fog
⚰️ 🐎
I really wish you could have a hand-free camera.
1:00:00
POST IN HD! Also, First.
RUclips just takes a while to process long videos, it's HD now
All I know is you need to do an IPO and value your company at $80 billion dollars. Do it now while the getting is good.
If Tesla FSD is to become a reality, this is the type of content I would expect to see from them....many years away it seems. Elon saying end of this year (once again)....I would wager otherwise.
2h30m of video and they never once opened the hood
The level of ML + SE required to solve this problem is insane.
The level of complexity brings to mind the Apollo space program.
This is inane
Putting a man on the moon is easier these days compared to FSD !
this explains Google maps and why they drove all over the world to capture street data.
If you designed your cities to be liveable, you would need to be inventing a car whose selling point is that you can spend that hour on an everyday way home from work sleeping.
What a dystopian society where the user story premise is a overworked pandemic ICU nurse whose afraid of her windows getting smashed in.
It was such a weird choice of narrative
Impressive tech for sure, and they've been able to get further than Tesla with the help of Lidar, but it's just a matter of time before they catch up given their scale, money and talent. Cruise is tied to a dead horse, GM could very well not survive the next 5 years. Either way they made 26 Chevy Bolts last quarter, while Telsa pumps out millions of cars every year, close to doubling that amount YoY. End of 2022 Tesla will have 3.7 million cars on the roads gathering data worldwide, and already has 60k "volunteer" safety-drivers who paid upwards of $10k for the privilege, probably close to a billion dollars now in (unrealized, yet eventually to be real) profit. I think Tesla FSD is going to be a lot worse for the foreseeable future, people will laugh at it, while slowly providing an increasingly compelling ADAS system to all kinds of driving, until one day they're suddenly there. At which point there will probably be 10+ million Teslas on the road, and so game over for pretty much everybody else.
I don't think this will beat NIO. NIO's technology will literally blow people's minds away.
I am blown away. Nio has the single pixel camera pedestrian detection. Booommmm!
Thank you so much for showing us how awful the boxes are, I would be Furious if all I got was less than a dollar for trading 500$ , they are making us look like fools … I’m just going to HODL my diamonds for now until they have more in the store. ❤️
Why not miniaturize the lidars into the windshield, headlights etc like cameras instead of elaborate stadium light arrangements that can never be mass commercialized? a laser and rocking little mirror are not big items.
Lidar may be optional, but radar is indispensable for AV/FSD Level 5.
Quite the opposite.
Lidar is useless, Lidar does not work when its raining, snowing etc. Tesla has the best approach for full self driving using a vision based system only.
@@solidn6 ah so camera works at night at snow at raining?
@@saltyfish157 Yes, watch Tesla AI day and Elons interview with Lex. ;)
@@solidn6 lmaoo sure
"Subjective" ☝️😏
Invent a pill that puts you to sleep for 30 minutes, gives you the quality & benefits of 8 hours- With no mental or physical degrading, No morning after issues aka 30 minute issues. Happy!
What would that do to the world? ouch.
Very techie. Not in my wheelhouse but OK
ok
Tesla is the answer.
Given how many billions they put into this waste of money of a project the outcome is modest at best... There thing does not work; even in their home town SF it blocks the road.
Kyle, just to let you know, if one of your bots kills someone I know - you are a dead man...
Keep in mind that cruise cars breaks down a lot in San Francisco. Just stops in the middle of the street for massively long times, blocking traffic for blocks, and in one case for miles here on highway 1 / 19th Avenue. Shit, the police cannot do anything about the cars when they malfunction.
So what your saying is that by having self driving cars being run by a computer that humans have programmed that humans can therefor be fully engaged on looking down and interacting on their mobile telephones 100% of the time? SMH (03:13)