Reduced vehicle “parked” attribute error rate by 17%, achieved by increasing the dataset size by 14%. Improved control for nearby obstacles by predicting continuous distance to static geometry with the general static obstacle network. Improved creeping profile with higher jerk when creeping starts and ends. This update improves autopilot control around fast-moving and cutting-in VRUs. To do this, we introduced a new dataset of simulated adversarial high-speed VRU interactions. Reduced the predicted velocity error of very close-by motorcycles, scooters, wheelchairs, and pedestrians by 63.6%. We find that this decreases the incidence of VRU-related false slowdowns. This was accomplished by increasing the data size of the next-gen autolabeler, training network parameters that were previously frozen, and modifying the network loss functions. Improved the precision of VRU detections by 44.9%, dramatically reducing spurious false-positive pedestrians and bicycles (especially around tar seams, skid marks, and raindrops). In particular, modeling intersection extents is now entirely based on network predictions and no longer uses map-based heuristics. Improved right-of-way understanding if the map is inaccurate or the car cannot follow the navigation. Use more accurate predictions of where vehicles are turning or merging to reduce unnecessary slowdowns for vehicles that will not cross our path. This enables us to predict crossing lanes, allows computationally cheaper and less error-prone post-processing, and paves the way for predicting many other signals and their relationships jointly and end-to-end. Upgraded modeling of lane geometry from dense rasters (“bag of points”) to an autoregressive decoder that directly predicts and connects “vector space” lanes point by point using a transformer neural network. 10.11 release notes, which should speak to what you see in the drone video, above: Here’s a quick rundown of Tesla’s FSD Beta v. Maybe will expand the Beta program w/ this release. Tesla was able to achieve these improvements by increasing the size of its next-generation labelers.įinally. Tesla noted that VRU detection improved by 44.9% in the newest update, and this allows its system to reduce “spurious false-positive pedestrians and bicycles.” This version of the software also revealed some key improvements, especially with the detection of vulnerable road users (VRU), which are typically pedestrians and cyclists. Tesla CEO Elon Musk recently tweeted that if FSD Beta version 10.11 performed well, Tesla could probably lower the minimum safety score to 95. The three-minute drone video below shows the trip as seen from above. Tesla Full Self Driving beta tester Whole Mars Catalog has shared a neat video of Tesla’s FSD Beta v.10.11 as seen from the sky.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |