Cruise AI under the Hood
0 likes326 views
The document discusses the technical components and challenges involved in the development of autonomous vehicles, including scene understanding, perception, and multiple sensor fusion. It highlights the importance of decision making, simulation, and continuous learning while addressing issues like uncertainty and occlusion. Additionally, it mentions the need for high-quality data extraction and the scalability of fleet management within the autonomous vehicle ecosystem.
1 of 61
Downloaded 13 times
![[DL輪読会]DIVERSE TRAJECTORY FORECASTING WITH DETERMINANTAL POINT PROCESSES](https://cdn.slidesharecdn.com/ss_thumbnails/kimura20200821dlseminarv1-200821021249-thumbnail.jpg?width=560&fit=bounds)
![[DL輪読会]画像を使ったSim2Realの現況](https://cdn.slidesharecdn.com/ss_thumbnails/imagesim2real-201030025320-thumbnail.jpg?width=560&fit=bounds)
Cruise AI under the Hood
- 1. Cruise.AI Under the Hood Yu Huang [email protected] Sunnyvvale, California
- 2. Outline • Scene Understanding • Perception • Multiple sensor fusion • Tracking • Occlusion • Prediction • Interaction/social modeling • Mixture of experts gating • Decision Making • Criterion & Challenges • Decision & Plan • Uncertainty & Interaction • Imitation & Reinforcement Learning • Latency handling • Simulation • Real2Sim • 3-D geometric modeling • Rare data generation • Lidar point cloud simulation • Ecosystem • Continuous learning • AV development cycle • High quality data extraction • Visualization tool • Huge data to leverage for ML models • Scalability of fleet management • HD map building at scale
- 20. Decision Making
- 38. Simulation