Research map / Video / Video segmentation and tracking
Learning-based Motion Planning in Dynamic Environments Using GNNs and Temporal Encoding
Ruipeng Zhang, Chenning Yu, Jingkai Chen, Chuchu Fan, Sicun Gao
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 DAgger training after behavior-cloning warm-up (ablation) | GNN-TE · 2Arms hard | success rate: 58.1±0.03 -> 62.5±0.02 | Improved |
| add DAgger training after behavior-cloning warm-up (ablation) | GNN-TE · 2Arms random | collision checking: 21.13±11.71 -> 17.24±8.45 | Improved |
| encode global obstacle trajectories in stage 1 (ablation) | GNN-TE · 2Arms hard | success rate | Improved |
| include local obstacle encoding over a time window in stage 2 (ablation) | GNN-TE · 2Arms | planning performance | Improved |
| add temporal encoding to both stages (ablation) | GNN-TE · 2Arms | planning performance | Improved |
| remove first-stage attention and temporal encoding and use only current-time obstacle encoding in stage 2 (ablation) | GNN-TE · 2Arms hard | success rate: 62.50% -> 32.00% | Hurt |
| add top-5 backtracking (ablation) | GNN-TE · 2Arms hard | success rate: 62.50% -> 89.30% | Improved |
| add top-5 backtracking (ablation) | GNN-TE · 2Arms random | success rate: 94.10% -> 98.00% | Improved |
| increase training problem count (ablation) | GNN-TE · 2Arms random and hard | success rate | Improved |
| train with extra hard examples (ablation) | GNN-TE · 2Arms hard | success rate: 62.5% -> 71.3% | Improved |
| use SIPP expert imitation with DAgger rather than end-to-end RL | GNN-TE vs DQN-GNN · 2Arms hard | success rate: 2.00% -> 62.50% | Improved |
| use GNN-TE temporal-spatial encoding and two-stage local planner instead of OracleNet-D obstacle concatenation | GNN-TE vs OracleNet-D · 2Arms hard | success rate: 10.80% -> 58.10% | Improved |
Closest papers by meaning
- Learning Neuro-Symbolic Abstractions for Robot Planning and Learning (AAAI 2024 · similarity 0.87)
- BaB-ND: Long-Horizon Motion Planning with Branch-and-Bound and Neural Dynamics (ICLR 2025 · similarity 0.86)
- Efficient Robotic Policy Learning via Latent Space Backward Planning (ICML 2025 · similarity 0.86)
- Physics-informed Temporal Difference Metric Learning for Robot Motion Planning (ICLR 2025 · similarity 0.86)
- Generative Trajectory Stitching through Diffusion Composition (NeurIPS 2025 · similarity 0.86)
- Hierarchical Planning and Learning for Robots in Stochastic Settings Using Zero-Shot Option Invention (AAAI 2024 · 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.
