Research map / Reinforcement learning / Visual and representation RL
Think2Drive: Efficient Reinforcement Learning by Thinking with Latent World Model for Autonomous Driving (in CARLA-v2)
Qifeng Li, Xiaosong Jia, Shaobo Wang, Junchi Yan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace model-free PPO with model-based RL in the expert (Think2Drive versus Roach-style PPO) | expert driving planner · CornerCaseRepo | driving score: 57.5 -> 83.8 | Improved |
| replace PPO with Think2Drive on official CARLA v2 routes | expert driving planner · CARLA Leaderboard v2 | driving score: 0.7 -> 56.8 | Improved |
| add reset technique (re-initialize planner parameters mid-training, Brick 1) (ablation) | Think2Drive · CornerCaseRepo training | success rate | Improved |
| add curriculum warm-up on simple lane following and turns (Brick 5) (ablation) | Think2Drive · CornerCaseRepo training | success rate | Improved |
| add termination-priority sampling (Brick 3) (ablation) | Think2Drive · CornerCaseRepo training | success rate | Improved |
| add steering cost function (Brick 4) (ablation) | Think2Drive · CornerCaseRepo training | success rate | Improved |
| increase planner train ratio to 4x the world model (Brick 6) (ablation) | Think2Drive · CornerCaseRepo training | success rate | Improved |
| distill Think2Drive expert into raw-sensor student (TCP) | TCP student · CornerCaseRepo | driving score: 83.8 -> 36.40 | Hurt |
| Think2Drive versus Roach on Leaderboard v1 official routes | expert driving planner · CARLA Leaderboard v1 | driving score: 84.0 -> 90.2 | Improved |
Later papers that compared with this method
1 later accepted paper reports Think2Drive as a baseline in their results tables; 0 of them report a better number on most of the results they share with it.
- Raw2Drive: Reinforcement Learning with Aligned World Models for End-to-End Autonomous Driving (in CARLA v2) (NeurIPS 2025 · better on 0 of 7 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Raw2Drive: Reinforcement Learning with Aligned World Models for End-to-End Autonomous Driving (in CARLA v2) (NeurIPS 2025 · similarity 0.89)
- Drive-R1: Bridging Reasoning and Planning in VLMs for Autonomous Driving with Reinforcement Learning (AAAI 2026 · similarity 0.88)
- DriveAgent-R1: Advancing VLM-based Autonomous Driving with Active Perception and Hybrid Thinking (ICLR 2026 · similarity 0.88)
- MAPLE: Latent Multi-Agent Play for End-to-End Autonomous Driving (NeurIPS 2026 · similarity 0.88)
- Beyond Self-Play and Scale: A Behavior Benchmark for Generalization in Autonomous Driving (NeurIPS 2026 · similarity 0.87)
- Bench2Drive: Towards Multi-Ability Benchmarking of Closed-Loop End-To-End Autonomous Driving (NeurIPS 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
