Roundabouts, solve the Chazman turn and navigate without GPS or map data

By Alex Jones

While unprotected left turns can be taxing for novice drivers, many FSD Beta participants still experience anxiety when their car is faced with such a turn.

Unprotected left turns are one of the most difficult driving maneuvers for humans and AI to perform, as they leave your vehicle vulnerable while turning (instead of turning with traffic).

Tesla quickly recognized the difficulties its AI had in executing these turns safely and worked diligently to improve the AI’s ability to discern impending hazards.

Solve Chazman’s tricky left turns

FSD Beta 10.12 improved unprotected left turn issues, but there are still lingering issues.

Chuck Cook (@Chazman) iteratively tested FSD Beta’s ability to make unprotected left turns on a busy road.

It’s a complicated turn that involved crossing three lanes of traffic before reaching the median and then turning left.

This often involves crossing all three lanes and waiting at the median before finding enough space to enter a lane.

It’s complicated, even for humans. However, Elon Musk said in a recent tweet that Beta 10.13 will fix Chuck’s complicated left turns “most of the time.”

Chazman on FSD beta 10.12.2

Smoother intersections

Elon also added that 10:13 would “smooth out intersection control”, hopefully reducing vehicle hesitation and steering jerks that can occur in busy intersections.

Go deep on roundabouts

In addition to unprotected left turns, navigating roundabouts and roundabouts has become a priority for FSD engineers.

Although popular overseas, roundabouts are rarely used in the United States. Although there are over 2600 roundabouts across the country, FSD Beta should be able to handle them well.

The roundabout navigation in FSD Beta was uneven, although it could be said that the multi-lane roundabouts are difficult even for experienced drivers.

With FSD Beta 10.13, Musk says Tesla is going “deep on roundabouts” and should be significantly better in this iteration of FSD Beta.

There are certainly tricky roundabouts that can challenge even the most experienced drivers. Some roundabouts include multiple lanes and may require you to change lanes inside the roundabout, some may even include roundabouts leading into or out of the roundabout, such as the “Magic Roundabout” in Swindon, in England.

Magic Roundabout in Swindon, England

Navigate without map data

Musk also hinted at another major improvement that could be included in 10.13: the ability for the FSD beta to navigate roads without map data.

The vehicle will be able to navigate to a specific GPS point or a pinned location (e.g. rural roads), Musk explains.

Dead reckoning (navigation without GPS)

Additionally, Elon hinted that Tesla is also working on AI’s ability to perform dead reckoning (navigation only on “inertial measurements, wheel motion, and vision”).

He gave underground car parks as an example of where FSD would need this capability to navigate without GPS or map data.

The car will be able to do this by using its last known GPS position and then determining its future position using only a compass, wheel movement and speed.

FSD Beta 10.13 Release Date

Elon said he expects FSD Beta 10.13 to be available in about two weeks.

Why is it always two weeks? Tesla can use an agile approach to software development. In agile development, work is planned in “sprints” that span one or more weeks, with two weeks being a common length.

Tesla may use two-week sprints when planning and developing upcoming features.

For drivers who have experienced the intense nature of FSD left turns through busy intersections, it looks like they won’t have to wait long for FSD Beta 10.13.

By Gabe Rodriguez Morrison

NHTSA updated its investigation involving Autopilot

Earlier this year, the National Highway Transportation Safety Administration (NHTSA) launched an investigation into Tesla’s “phantom braking” issue, which occurs when the vehicle suddenly slows down.

In February, NHTSA said “rapid deceleration can occur without warning, randomly, and often repeatedly during a single drive cycle.” The problem is of particular concern on highways where hard braking could lead to a rear-end collision.

Since then, NHTSA has officially turned its preliminary investigation into a “technical analysis,” which is conducted before the agency decides on a recall. Regulators will assess 830,000 vehicles after several cars from the automaker collided with stopped emergency vehicles while Autopilot was engaged.

Tesla’s Autopilot feature is designed to help drivers navigate the roads by using a combination of cameras and artificial intelligence to detect other vehicles, pedestrians, stop lights, and more. Tesla asks drivers to pay attention to the road and keep their hands on the wheel when using Autopilot.

In the past four years, 16 Teslas with Autopilot have rammed into parked first aid vehicles, injuring 15 and killing one. According to forensic data, most drivers had their hands on the steering wheel before impact, per Tesla’s instructions.

In this investigation, regulators will determine whether or not Tesla’s Autopilot feature compromises “effective driver supervision.”

“The investigation will assess the technologies and methods used to monitor, assist and enforce driver engagement in the dynamic driving task during Autopilot operation,” NHTSA said.

This survey covers approximately 830,000 Tesla vehicles that were built after 2014. Tesla launched its Autopilot feature that year and then moved to its own Autopilot hardware in 2016 after ending its partnership with MobileEye.

Tesla has yet to respond to the extensive investigation into its Autopilot feature.

2 days since announcement

By Alex Jones

Elon says more animal visualizations are coming

Lions, tigers and bears, oh my! Although Dorothy’s list of animals isn’t included in Elon Musk’s latest tweet about Tesla’s neural network recognizing animals, dogs, cats, and horses are included.

As for Tesla’s object recognition capabilities, it looks like Tesla is preparing to release an update that includes new visualizations of various animals that may stray into the car’s path.

Regarding animal recognition, Teslas currently only displays dogs, but it will often display a dog even if it is another animal.

This is due to a lack of training, where the car cannot yet tell the difference between a dog, a horse or a cat. It can be easily solved by training their NN by adding and categorizing additional animal photos.

Looks like we’ll have some new animal visualizations soon. The most obvious may be animals that are regularly found near the streets, such as deer, horses, possums, or other creatures.

While automotive competitors have made strides in developing electric vehicles after years behind Tesla, there’s no denying that Tesla’s neural networks are state-of-the-art.

These neural networks allow Tesla to collect live data from more than one million participating vehicles. They use these large datasets of real-world data to train their AI algorithms to identify objects that may pose a hazard to drivers.

Tesla claims that “a full Autopilot neural network build involves 48 networks that require 70,000 GPU hours to train.”

While users have noticed drastic improvements in the car’s ability to recognize different types of cars (as seen in improved visualizations in 2022.16), the likely addition of new animals is a nice addition.

With FSD Beta 10.12, Tesla has added many visualization updates. Will more animals be added in FSD Beta 10.13, which is expected in about two weeks?

While it’s too early to tell, it’s still exciting that Teslas’ objection recognition abilities continue to improve.

Comments are closed.