Tesla has continued with its 2-weekly cadence of releases of its Full Self-Driving beta software, releasing FSDBeta v9.2 in the early hours of this morning.
Telsa has called the update a minor release, but the changelog appears relatively beefy:
- Clear-to-go boost through turns on minor-to-major roads (plan to expand to all roads V9.3)
- Improved peek behaviour where we are smarter about when to go around the lead vehicle by reasoning about the causes for lead vehicle being slow
- v1 of the Multi-modal predictions for where other vehicles expected to drive. This is only partially consumed for now.
- New Lanes network with 50 more clips (almost double) from the new auto-labelling pipeline
- New VRU velocity model with 12% improvement to velocity and better VRU clear-to-go performance. This is the first model trained with “Quantization-Aware-Training”, an improved technique to mitigate int8 quantization.
- Enabled Inter-SoC synchronous compute scheduling between vision and vector space processes. Planner in the loop is happening in V10.
- Shadow mode for new crossing/merging targets network which will help improve VRU control
Chuck Cook has posted a video showing the software in action.
The clearest improvement appears to be a new acceleration behaviour, where the car would speed up to take advantage of opportunities, much like human drivers would.
As usual, it seems clear there is still a lot of work to do. Consistent with this, Elon Musk has now said there will be no wide release until version 10, of of course it is not clear when this will arrive.
What do our readers think of the performance of the latest version of Tesla’s Full Self-driving software? Let us know in the comments below.