Tesla begins rolling out big full self-driving beta update

Tesla has started rolling out a big new full self-driving (FSD) beta software update with lots of improvements.

According to CEO Elon Musk, the automaker could expand access to the beta if the update “performs well.”

Tesla full self-driving beta

Starting in October 2020, Tesla is gradually rolling out what it calls “Full Self-Driving Beta” (FSD Beta), an early version of its self-driving software that is currently being tested by a fleet of Tesla owners. Through the company and its “Safety Test Score”.

The software enables the vehicle to drive autonomously to the destination entered in the car’s navigation system, but requires the driver to be alert and ready to take control at all times.

Because the responsibility lies with the driver and not Tesla’s system, it is still considered a Level Two driver-assist system, despite its name. It has been a “two steps forward, one step back” type of programme, as some of the updates have seen regression in terms of driving capabilities.

Tesla has been releasing new software updates to the FSD beta program frequently and adding more owners to it.

The last significant update was FSD beta 10.10 in early February.

As of Q4 2021, the automaker said it had about 60,000 owners in the FSD beta program.

Tesla FSD Beta 10.11

Now Tesla has started rolling out a new FSD beta 10.11 update to its Early Access program, and it’s a significant one, based on the release notes.

Here are the release notes:

  • Upgraded from a dense raster (“bag of points”) of lane geometry to an autoregressive decoder that directly predicts and adds “vector space” lane point by point using a transformer neural network. This enables us to predict crossing lanes, allows computationally cheaper and less error-prone post-processing, and a way to predict many other signals and their relationships jointly and end-to-end. paves. Use more accurate predictions of where vehicles are turning or merging to reduce unnecessary deceleration for vehicles not crossing our path.
  • Better understanding if the map is wrong or the car can’t follow the navigation. Notably, modeling intersection extension is now based entirely on network predictions and no longer uses map-based estimation.
  • Improved the accuracy of VRU detection by up to 44.9%, dramatically reducing false positives of pedestrians and bicycles (especially around tar seams, skid marks and raindrops). This was accomplished by increasing the data size of the next generation autolabeler, training the previously frozen network parameters, and modifying the network loss functions. We find that this leads to a reduction in the incidence of false recessions related to VRUs.
  • Reduced estimated velocity error of motorcycles, scooters, wheelchairs and pedestrians by 63.6%. To do this, we introduced a new dataset of simulated aversive high-speed VRU interactions. This update also improves autopilot controls around the VRU for faster running and cutting.
  • Improved creep profile with higher shocks when creep starts and ends.
  • Better control for nearby obstacles by predicting a constant distance from the stationary geometry with a normal static constraint network.
  • A 17% reduction in the vehicle “park” attribute error rate, achieved by increasing the dataset size by 14%. Brake light accuracy also improved.
  • Improved clear-to-go landscape velocity error 5% and highway landscape velocity error 10%, achieved by a tuning loss function targeted to improve performance in difficult scenarios.
  • Improved detection and control for open car doors.
  • Improved smoothness through turns using an optimization-based approach to decide which road lines are irrelevant to lateral and longitudinal acceleration and jerk limits as well as controlling vehicle kinematics .
  • Improved stability of FSD UL visualization by optimizing the Ethernet data transfer pipeline by up to 15%.

CEO Elon Musk says that if this update “performs well,” Tesla will “probably” reduce access to drivers who score ’95’ on the Driver Safety Score. This should give a lot of FSD buyers access to the beta.

As for Canada, Tesla owners who bought FSDs have had access to driver safety scores for more than a week now. The automaker is expected to start releasing betas to those with the top scores soon.

FTC: We use income generating auto affiliate links. More.

Subscribe to Electrek on YouTube for exclusive videos and subscribe to podcasts.

Comments

Post a Comment

Popular posts from this blog

Tesla Model Y welcomed by Kansas PD, expected to give cost savings – Elon Musk - Elon Musk

In a recent press release, the Leawood Police Department in Kansas announced that it had acquired a Tesla Model Y Long Range for its fleet. The vehicle would be used as one of the Leawood PD’s patrol cars. The Model Y has already been painted in the Leawood PD’s black and white colors. The department noted that they started looking into the idea of an all-electric patrol car last year after seeing dismal reviews of available hybrid patrol cars. Captain Kirt Yoder, the agency’s project manager, noted that Tesla Kansas had been quite helpful. “I ended up contacting about ten agencies from New York to California. Tesla-Kansas City was great. They were eager to help and answer all our questions,” he said. Apart from being a green vehicle, the Model Y is also expected to give the Leawood Police Department substantial cost savings. Over a year ago, a cost comparison already showed that the department’s Ford Explorer patrol SUV would cost about $4,500 per year to run. That was a time w...
Read more

Tesla Autopilot workers confirm mass layoff in San Mateo office – Elon Musk - Elon Musk

Image
Tesla’s mass layoffs reached the Autopilot team in San Mateo. Autopilot workers have started posting the news on LinkedIn.  People familiar with the matter told  Bloomberg  that Tesla laid off about 200 Autopilot workers from its San Mateo office. San Mateo had around 350 employees before the recent layoffs. Tesla transferred some of the San Mateo staff to a nearby facility in previous weeks.  The Autopilot team in San Mateo worked on customer vehicle data, specifically data labeling, for Tesla’s autonomous driver assistance features. According to a source, most of the employees Tesla laid off were hourly workers from the Autopilot Team. Most of the San Mateo office staff were data annotation specialists who held hourly positions.  One data annotation specialist, Caesar Rosas, shared that he received the news of Tesla’s layoffs recently. He stated that almost everyone working at Tesla’s San Mateo branch was laid off, adding that he didn’t get to s...
Read more

Sen. Joe Manchin blasts EV tax credit expansion, calls for hydrogen development

West Virginia Democrat Joe Manchin blasted a potential expansion of the electric vehicle (EV) tax credit on Thursday, while also calling for the development of the hydrogen sector to decarbonize the transportation industry. Manchin, who called a Biden administration proposal to broaden the EV tax credit program “ludicrous,” said the waiting period for EVs right now is already so long that there is no point to offering more money to incentivize consumers to buy a battery-powered car. “There is a waiting list for EVs right now, with the fuel prices, but they still want us to throw a $5,000, or $7,000 or $12,000 credit for us to buy electric vehicles,” said during a Thursday Senate budget hearing. Manchin added that production volumes of EVs are far away from fulfilling current demand levels, so it would be ridiculous to offer additional funding to the programs. It is evident that EVs are becoming more popular, and with automakers in nearly all situations relating to EVs still workin...
Read more