Research map / Reinforcement learning / Visual and representation RL
Unlocking Efficient Vehicle Dynamics Modeling via Analytic World Models
Asen Nachkov, Danda Pani Paudel, Jan-Nico Zaech, Davide Scaramuzza, Luc Van Gool
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 |
|---|---|---|---|
| train the policy with improved APG (winner-take-all mixture etc.) instead of previous APG | APG policy · Waymax route-conditioned | ADE: 2.0083 -> 1.8121 | Improved |
| train the odometry world model through the differentiable simulator (ablation) | Odometry AWM · in-distribution ADE at 1.5 sec | ADE [m]: 1.6200 -> 0.5496 | Improved |
| train planner with DiffSim and inverse kinematics instead of previous APG policy | Planner AWM · Waymax route-conditioned | ADE: 2.0083 -> 1.8734 | Improved |
| sample more rollouts without route conditioning (ablation) | APG policy · Waymax no route conditioning | min ADE: 3.5725 -> 1.1414 | Improved |
| select actions by inverse-state norm instead of distance to next log state (ablation) | MPC · Waymax | ADE: 1.8136 -> 1.8138 | No clear effect |
| add MPC with 8 imagined rollouts over 10 steps versus reactive single rollout (ablation) | AWM agent · Waymax | ADE: 3.5883 -> 3.4719 | Improved |
| lengthen MPC imagination from 10 to 20 steps (ablation) | MPC with 8 rollouts · Waymax | ADE: 3.4719 -> 3.2179 | Improved |
| increase MPC rollouts from 8 to 12 (top-4) (ablation) | MPC · Waymax | ADE: 3.2179 -> 3.2486 (20 steps) | No clear effect |
Closest papers by meaning
- AdaWM: Adaptive World Model based Planning for Autonomous Driving (ICLR 2025 · similarity 0.87)
- Driving in the Occupancy World: Vision-Centric 4D Occupancy Forecasting and Planning via World Models for Autonomous Driving (AAAI 2025 · similarity 0.87)
- Vista: A Generalizable Driving World Model with High Fidelity and Versatile Controllability (NeurIPS 2024 · similarity 0.87)
- Diffusion-Based Planning for Autonomous Driving with Flexible Guidance (ICLR 2025 · similarity 0.86)
- Future-Aware End-to-End Driving: Bidirectional Modeling of Trajectory Planning and Scene Evolution (NeurIPS 2025 · similarity 0.86)
- UniMotion: A Unified Motion Framework for Simulation, Prediction and Planning (NeurIPS 2025 · similarity 0.86)
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.
