Research map / Reinforcement learning / Policy gradient and actor-critic
Implicit Action Chunking for Smooth Continuous Control
Bosun Liang, Shuo Pei, Zirui Chen, Chuanzhi Fan, Chen Sun, Yuankai Wu, Huachun Tan, Yong Wang
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 |
|---|---|---|---|
| add execution window to HG-TD3 (ablation) | HG-TD3 · CARLA LCO | success rate: 0.00 -> 0.50 | Improved |
| add value window to HG-TD3 (ablation) | HG-TD3 · CARLA LCO | success rate: 0.00 -> 0.65 | Improved |
| add smoothness regularizer to HG-TD3 (ablation) | HG-TD3 · CARLA LCO | success rate: 0.00 -> 0.60 | Improved |
| add regularizer to value window (ablation) | Value Window · CARLA LCO | success rate: 0.65 -> 0.80 | Improved |
| add execution window to value window plus regularizer (ablation) | Value Window + Reg · CARLA LCO | success rate: 0.80 -> 1.00 | Improved |
| add execution window to value window plus regularizer (ablation) | Value Window + Reg · CARLA LCO | AbsAccP95: 1.381 -> 1.195 | Improved |
| use zero-order hold instead of dissipative decay (ablation) | DWS · CARLA AEB | reward: 388.66 -> 390.70 | Improved |
| use zero-order hold instead of dissipative decay (ablation) | DWS · CARLA AEB | JerkRMS_a: 49.24 -> 47.90 | Improved |
| increase window size h from 1 to 3 (ablation) | DWS · EMS | reward: -2179 -> -1393 | Improved |
| increase window size h from 3 to 5 (ablation) | DWS · EMS | reward: -1393 -> -2387 | Hurt |
| add smoothness regularization lambda_S=0.10 vs none (ablation) | DWS · EMS | reward: -2355 -> -1393 | Improved |
| raise lambda_S from 0.10 to 0.40 (ablation) | DWS · EMS | reward: -1393 -> -2849 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- LipsNet++ — LipsNet++: Unifying Filter and Controller into a Policy Network (ICML 2025) · better on 6 of 6 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
- Enhancing Control Policy Smoothness by Aligning Actions with Predictions from Preceding States (AAAI 2026 · similarity 0.88)
- Real-Time Execution of Action Chunking Flow Policies (NeurIPS 2025 · similarity 0.87)
- Reinforcement Learning with Action Chunking (NeurIPS 2025 · similarity 0.84)
- PACE: Phase-Aware Chunk Execution for Robot Policies with Action Chunking (NeurIPS 2026 · similarity 0.84)
- FASTER: Rethinking Real-Time Flow VLAs (NeurIPS 2026 · similarity 0.84)
- SEAR: Sample Efficient Action Chunking Reinforcement Learning (NeurIPS 2026 · similarity 0.84)
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.
