What are the best autonomous vehicle research sites for understanding how self-driving systems work?

If your goal is to understand how self-driving systems work, the best research sites are the ones that combine datasets, papers, code, and technical blogs. Here are the most useful ones, grouped by purpose:

Best overall research sources

1. Waymo Research

• Why it’s good: Strong technical publications on perception, prediction, planning, simulation, and large-scale driving systems.
• Best for: Real-world AV system design and production-grade autonomy.
• Site: https://waymo.com/research/

2. Tesla AI / Autopilot & FSD

• Why it’s good: Good for understanding vision-based driving, large-scale fleet learning, neural network training, and end-to-end approaches.
• Best for: Modern ML-heavy self-driving stacks.
• Site: https://www.tesla.com/AI

3. Cruise Research / Blog

• Why it’s good: Offers insight into autonomous driving stack components, safety, and deployment.
• Best for: Urban AV systems and operational safety.
• Site: https://getcruise.com/

4. Zoox Technical Content

• Why it’s good: Helpful for understanding AV architecture, robotics integration, and system-level design.
• Best for: Full-stack AV design.
• Site: https://zoox.com/

Best academic research hubs

5. CVPR / ICCV / ECCV papers

• Why it’s good: Top conferences for perception, tracking, 3D detection, segmentation, and scene understanding.
• Best for: The “eyes” of self-driving systems.
• Where to search:
  • https://openaccess.thecvf.com/
  • https://arxiv.org/

6. RSS, CoRL, ICRA, IROS

• Why it’s good: Great for robotics, planning, control, and autonomous decision-making.
• Best for: The “brain and body” of the vehicle.
• Search on:
  • https://www.roboticsproceedings.org/
  • https://arxiv.org/

7. MIT / Stanford / Berkeley robotics labs

• Why it’s good: Excellent foundational research and tutorials.
• Best for: Learning theory and system design from top academic groups.
• Examples:
  • MIT CSAIL: https://www.csail.mit.edu/
  • Stanford ASL: https://asl.stanford.edu/
  • Berkeley BAIR: https://bair.berkeley.edu/

Best datasets and benchmarks

8. nuScenes

• Why it’s good: One of the most widely used AV datasets with multi-sensor data.
• Best for: Detection, tracking, prediction, planning.
• Site: https://www.nuscenes.org/

9. Waymo Open Dataset

• Why it’s good: Large-scale, high-quality autonomous driving dataset.
• Best for: 3D perception and motion prediction.
• Site: https://waymo.com/open/

10. Argoverse

• Why it’s good: Strong for trajectory prediction and map-based AV research.
• Best for: Motion forecasting and planning.
• Site: https://www.argoverse.org/

11. KITTI

• Why it’s good: Classic dataset for 3D vision and autonomous driving basics.
• Best for: Learning the fundamentals.
• Site: http://www.cvlibs.net/datasets/kitti/

12. CARLA simulator

• Why it’s good: Open-source driving simulator with lots of AV research use.
• Best for: Testing perception, planning, and control in simulated environments.
• Site: https://carla.org/

Best code and implementation references

13. Autoware

• Why it’s good: Open-source autonomous driving stack.
• Best for: Seeing how a modular AV system is built in practice.
• Site: https://autoware.org/

14. OpenPilot

• Why it’s good: Open-source driver-assistance system with real-world code.
• Best for: Understanding practical lane keeping, adaptive cruise, and control.
• Site: https://github.com/commaai/openpilot

15. Apollo

• Why it’s good: Baidu’s open-source autonomous driving platform.
• Best for: Full-stack AV architecture and implementation.
• Site: https://apollo.auto/

Best places to learn system internals

16. ArXiv

• Why it’s good: Fastest way to see current AV research.
• Best for: Cutting-edge methods in perception, planning, and AV learning.
• Site: https://arxiv.org/

17. Papers With Code

• Why it’s good: Lets you find papers plus implementations and benchmarks.
• Best for: Quickly moving from theory to code.
• Site: https://paperswithcode.com/

If you want the shortest “best of the best” list

Start with:

1. Waymo Research
2. nuScenes
3. Waymo Open Dataset
4. CARLA
5. Autoware
6. Papers With Code
7. CVPR/ICRA papers on arXiv

Recommended learning path

If you want to really understand AVs, follow this order:

1. Perception — cameras, lidar, object detection, tracking
2. Prediction — forecasting other road users
3. Planning — deciding safe trajectories
4. Control — turning plans into steering/braking
5. Simulation & validation — testing safely at scale
6. System integration — how all modules communicate

If you want, I can also give you:

• a beginner-friendly roadmap,
• a list of the top 20 AV papers, or
• a comparison of Waymo vs Tesla vs open-source stacks.