Research map / Vision-language and multimodal / Robot action and navigation
SpikePingpong: Spike Vision-based Fast-Slow Pingpong Robot System
Hao Wang, Chengkai Hou, Xianglong Li, Yankai Fu, Chenxuan Li, Ning Chen, Gaole Dai, Jiaming Liu, Tiejun Huang, Shanghang Zhang
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 System 2 neural calibration to System 1 (ablation) | SpikePingpong · Ball hittable position prediction | overall MAE: 44.13 -> 12.34 | Improved |
| use System 2 instead of an RNN-based predictor (ablation) | SpikePingpong · Ball hittable position prediction | overall MAE: 22.80 -> 12.34 | Improved |
| add System 2 calibration to System 1 with IMPACT (ablation) | SpikePingpong · Single-target return (30cm) | overall success rate: 23 -> 92 | Improved |
| use Fast-Slow perception instead of RNN with IMPACT (ablation) | SpikePingpong · Single-target return (30cm) | overall success rate: 67 -> 92 | Improved |
| add System 2 calibration to System 1 with IMPACT (ablation) | SpikePingpong · Sequential targets (30cm) | overall success rate: 15 -> 78 | Improved |
| use Fast-Slow perception instead of RNN with IMPACT (ablation) | SpikePingpong · Sequential targets (30cm) | overall success rate: 52 -> 78 | Improved |
| remove vision input from ACT | ACT · Single-target return (30cm) | average success rate: 12 -> 19 | Improved |
| remove vision input from Diffusion Policy | Diffusion Policy · Single-target return (30cm) | average success rate: 3 -> 6 | Improved |
| move launcher to unseen off-center positions (OOD) | SpikePingpong · Single-target return | 30cm success rate: 92 -> 74 | Hurt |
| move launcher to unseen off-center positions (OOD) | SpikePingpong · Single-target return | 20cm success rate: 70 -> 52 | Hurt |
| test on an unseen human player instead of the fine-tuned one | SpikePingpong · Human demonstration fine-tuning (30cm) | average success rate: 47 -> 31 | Hurt |
| test on an unseen human player instead of the fine-tuned one | SpikePingpong · Human demonstration fine-tuning (20cm) | average success rate: 27 -> 15 | Hurt |
Closest papers by meaning
- LATTE-MV: Learning to Anticipate Table Tennis Hits from Monocular Videos (CVPR 2025 · similarity 0.86)
- AutoSpeed: Annotation-Free Stage-Adaptive Motion Speed Learning for Robot Manipulation (ECCV 2026 · similarity 0.85)
- Chain-of-Action: Trajectory Autoregressive Modeling for Robotic Manipulation (NeurIPS 2025 · similarity 0.84)
- DexTrack: Towards Generalizable Neural Tracking Control for Dexterous Manipulation from Human References (ICLR 2025 · similarity 0.84)
- SAM2Act: Integrating Visual Foundation Model with A Memory Architecture for Robotic Manipulation (ICML 2025 · similarity 0.84)
- Predictive Inverse Dynamics Models are Scalable Learners for Robotic Manipulation (ICLR 2025 · 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.
