Research map / Reinforcement learning / Imitation learning
TarGF: Learning Target Gradient Field to Rearrange Objects without Explicit Goal Specification
Mingdong Wu, fangwei zhong, Yulong Xia, Hao Dong
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 TarGF with ORCA planning instead of the planning-based and learning-based baselines | Ball Rearrangement · Circling, Clustering, and Circling+Clustering | pseudo-likelihood | Improved |
| Use TarGF with ORCA planning instead of the baselines | Ball Rearrangement · Circling, Clustering, and Circling+Clustering | coverage score | Improved |
| Use TarGF with SAC instead of the strongest goal-based baseline | Ball Rearrangement · Clustering and Circling+Clustering | coverage score | Hurt |
| Use TarGF-based control instead of learning-based and planning-based baselines | Ball Rearrangement · Circling, Clustering, and Circling+Clustering | absolute state change | Improved |
| Use TarGF-guided SAC instead of learning-based baselines | Room Rearrangement · 3D-FRONT test rooms | coverage score | Improved |
| Use TarGF-guided SAC instead of learning-based baselines | Room Rearrangement · 3D-FRONT test rooms | averaged collision number | Improved |
| Add residual policy learning to the TarGF-guided policy (ablation) | Ours (SAC) · Ball Rearrangement | coverage score | Improved |
| Add a control algorithm to adapt gradient-based actions to environment dynamics (ablation) | TarGF-based control · Ball Rearrangement | pseudo-likelihood | Improved |
| Add a control algorithm to adapt gradient-based actions to environment dynamics (ablation) | TarGF-based control · Ball Rearrangement | averaged collision number | Improved |
| Zero-shot transfer the trained policy to environments with more balls | TarGF-based methods · Ball Rearrangement with increased ball counts | pseudo-likelihood increment | Improved |
Closest papers by meaning
- Task Planning for Object Rearrangement in Multi-Room Environments (AAAI 2024 · similarity 0.87)
- Entity-Centric Reinforcement Learning for Object Manipulation from Pixels (ICLR 2024 · similarity 0.85)
- DiffTORI: Differentiable Trajectory Optimization for Deep Reinforcement and Imitation Learning (NeurIPS 2024 · similarity 0.85)
- ManiSkill-HAB: A Benchmark for Low-Level Manipulation in Home Rearrangement Tasks (ICLR 2025 · similarity 0.84)
- Efficient Exploration and Discriminative World Model Learning with an Object-Centric Abstraction (ICLR 2025 · similarity 0.84)
- GO-DICE: Goal-Conditioned Option-Aware Offline Imitation Learning via Stationary Distribution Correction Estimation (AAAI 2024 · 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.
