Research map / Architectures and efficiency / RNNs and normalization
PyPose: A Library for Robot Learning With Physics-Based Optimization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 PyPose for dense Lie group Jacobian computation | Lie group operations · Dense Jacobian, CPU | speedup: up to 37.9× faster than Theseus; up to 8.8× faster than LieTorch | Helped |
| Use PyPose for dense Lie group Jacobian computation | Lie group operations · Dense Jacobian, GPU | speedup: up to 42.9× faster than Theseus; up to 8.6× faster than LieTorch | Helped |
| Use PyPose for batched Lie group operations and Jacobians | Lie group operations · Batched operations and Jacobians, CPU and GPU | speedup: CPU operations: up to 1.8×, 2.8×, and 2.2×; GPU operations: up to 7.4×, 32.9×, and 17.4×; CPU Jacobians: up to 7.3×, 16. | Helped |
| Use second-order LM instead of SGD | PyPose transform-inverse optimization · Transform inverse, batches 10^1–10^4 | final error: 5.10E+1 -> 5.96E-5 (batch 10^1); 4.96E+2 -> 5.80E-4 (batch 10^2); 4.09E+3 -> 3.72E-3 (batch 10^3); 6.86E+3 -> 3.77E-2 (b | Helped |
| Use second-order LM instead of Adam | PyPose transform-inverse optimization · Transform inverse, batches 10^1–10^4 | final error: 9.81E+1 -> 5.96E-5 (batch 10^1); 1.25E+3 -> 5.80E-4 (batch 10^2); 1.13E+4 -> 3.72E-3 (batch 10^3); 1.16E+5 -> 3.77E-2 (b | Helped |
| Add self-supervised bundle-adjustment training to CAPS | CAPS feature matcher · TartanAir Easy | mean matching accuracy: 21.37% -> 32.04%; + 49.94% | Helped |
| Add self-supervised bundle-adjustment training to CAPS | CAPS feature matcher · TartanAir Hard | mean matching accuracy: 18.10% -> 24.50%; + 35.33% | Helped |
| Add self-supervised bundle-adjustment training to CAPS | CAPS feature matcher · TartanAir Overall | mean matching accuracy: 19.74% -> 28.27%; + 43.24% | Helped |
| Train an IMU calibration network and differentiate through IMU preintegration | IMU preintegration · EuRoC MH_02_easy | position RMSE: 0.050 -> 0.012 | Helped |
| Train an IMU calibration network and differentiate through IMU preintegration | IMU preintegration · EuRoC MH_02_easy | rotation RMSE: 0.659 -> 0.012 | Helped |
| Train an IMU calibration network and differentiate through IMU preintegration | IMU preintegration · EuRoC MH_04_hard | position RMSE: 0.047 -> 0.017 | Helped |
| Train an IMU calibration network and differentiate through IMU preintegration | IMU preintegration · EuRoC MH_04_hard | rotation RMSE: 0.652 -> 0.016 | Helped |
Closest papers by meaning
- Deep Reinforcement Learning for Robotics: A Survey of Real-World Successes (AAAI 2025 · similarity 0.85)
- Predictive Inverse Dynamics Models are Scalable Learners for Robotic Manipulation (ICLR 2025 · similarity 0.84)
- Scaling Proprioceptive-Visual Learning with Heterogeneous Pre-trained Transformers (NeurIPS 2024 · similarity 0.84)
- Rapidly Adapting Policies to the Real-World via Simulation-Guided Fine-Tuning (ICLR 2025 · similarity 0.84)
- SAM2Act: Integrating Visual Foundation Model with A Memory Architecture for Robotic Manipulation (ICML 2025 · similarity 0.84)
- UniDomain: Pretraining a Unified PDDL Domain from Real-World Demonstrations for Generalizable Robot Task Planning (NeurIPS 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.