Research map / Reinforcement learning / Visual and representation RL
Neural Dynamic Policies for End-to-End Sensorimotor Learning
Shikhar Bahl, Mustafa Mukadam, Abhinav Gupta, Deepak Pathak
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 |
|---|---|---|---|
| Use NDP instead of NN | behavior cloning · Throwing | success rate: 0.528 $\pm$ 0.262 -> 0.642 $\pm$ 0.246 | Improved |
| Use NDP instead of NN | behavior cloning · Picking | success rate: 0.672 $\pm$ 0.074 -> 0.408 $\pm$ 0.058 | Hurt |
| Use NDP instead of NN | behavior cloning · Pushing | success rate: 0.002 $\pm$ 0.004 -> 0.208 $\pm$ 0.049 | Improved |
| Use NDP instead of NN | behavior cloning · Soccer | success rate: 0.885 $\pm$ 0.016 -> 0.890 $\pm$ 0.010 | No clear effect |
| Use NDP instead of NN | behavior cloning · Faucet-Opening | success rate: 0.532 $\pm$ 0.231 -> 0.790 $\pm$ 0.059 | Improved |
| Predict dynamical-system parameters with NDP rather than a CNN | digit-writing behavior cloning · Digit writing train | mean trajectory loss: 10.42 $\pm$ 5.26 -> 0.70 $\pm$ 0.36 | Improved |
| Predict dynamical-system parameters with NDP rather than CNN-DMP | digit-writing behavior cloning · Digit writing held-out test | mean trajectory loss: 8.46 $\pm$ 8.45 -> 0.74 $\pm$ 0.34 | Improved |
| Remove the forcing term and learn only the goal g (ablation) | NDP · Pushing | sample efficiency | Hurt |
| Remove the forcing term and learn only the goal g (ablation) | NDP · Pushing | asymptotic performance | Hurt |
| Learn global α instead of keeping it fixed (ablation) | NDP · Pushing | performance | Hurt |
| Reduce rollout length to k=3 (ablation) | NDP · Pushing | performance | Hurt |
| Vary the number of basis functions, integration steps, and basis-function type (ablation) | NDP · Pushing | performance | No clear effect |
Closest papers by meaning
- Latent Policy Barrier: Learning Robust Visuomotor Policies by Staying In-Distribution (NeurIPS 2025 · similarity 0.89)
- Make-An-Agent: A Generalizable Policy Network Generator with Behavior-Prompted Diffusion (NeurIPS 2024 · similarity 0.88)
- Scaling Proprioceptive-Visual Learning with Heterogeneous Pre-trained Transformers (NeurIPS 2024 · similarity 0.88)
- DynaMo: In-Domain Dynamics Pretraining for Visuo-Motor Control (NeurIPS 2024 · similarity 0.88)
- Contractive Dynamical Imitation Policies for Efficient Out-of-Sample Recovery (ICLR 2025 · similarity 0.88)
- Act to See, See to Act: Diffusion-Driven Perception-Action Interplay for Adaptive Policies (NeurIPS 2025 · 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.
