Added new Vision Speed network to infer the typical driving speed on a given road. Improved speed control during lane changes through better consideration of upcoming navigation deadlines, required back-to-back lane changes and presence of a vehicle behind ego. Improved understanding for when to use bus lanes and when to avoid them, by updating the lane type detection network and improving map-vision fusion. This was accomplished by further increasing our auto-labeled fleet dataset by 80k clips, improving the accuracy of the auto-labeling algorithm, and tuning the distribution of training supervision. Improved recall for partial cut-ins by 39% and precision for false positive cut-ins due to lane changes into adjacent lanes by 66%, resulting in a 33% reduction in overall lane-changing prediction error. Improved geometric consistency between lane, line, road edge and restricted space detections by re-training our networks on the same dataset with the latest version of our "lane guidance" module, and by using a common features space to predict line, road edge and restricted space. These changes combined resolved 64% of all interventions caused by bad routing type. Also significantly improved per-lane routing type autolabeler. Improved lane guidance module to feed in long range routing "hints" to the network for which lanes ego needs to be in to reach its destination. Examples of improved cases include: turning when the turn lane is blocked by parked cars and avoiding turning into bus lanes. Improved turn performance in dense unstructured city environments. Improved ego's behavior near VRUs by measuring their probability of intersecting ego's path, based on their kinematic data, and preemptively decelerating when the estimated risk is high. Improved the decision to assert or yield for pedestrians at more crosswalks by evaluating multiple possible futures in the joint space of ego's actions and the pedestrian's response.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |