A field guide to the robot manipulation literature

What it does

This repository is a curated bibliography and taxonomy of robot manipulation research. It sorts papers into categories ranging from grasping and dexterous manipulation to high-level task planning and low-level policy learning, including diffusion policies and vision-language-action models. The maintainers also link to code repositories when available and track simulators, benchmarks, and datasets alongside the literature.

The interesting bit

Rather than dumping links, the repository builds a hierarchical classification system that tries to map the entire field—from aerial and underwater manipulation to bimanual grasping and soft robotics—into a single navigable structure. It also backs a companion survey paper, so the curation doubles as the reading list for a formal literature review.

Key highlights

• Covers both classical non-learning control and modern learning-based methods (imitation learning, RL, VLAs)
• Tracks specialized subfields often ignored in general ML lists: soft robotic manipulation, deformable objects, quadrupedal and humanoid manipulation
• Includes sections on practical bottlenecks like data collection, generalization, and cross-embodiment transfer
• Maintains lists of simulators, benchmarks, and datasets alongside the papers
• Actively revised: the taxonomy itself gets refactored as the field shifts (e.g., recent splits for video-based planners and aerial manipulation)

Caveats

• The README is largely a table of contents; the detailed paper lists live in linked markdown files, so browsing requires some clicking around
• Code availability is inconsistent; the maintainers recently pruned entries lacking publicly available repositories, but many papers still have no linked code
• Breadth comes at the cost of depth in any single niche, and the maintainers explicitly welcome community contributions to fill gaps

Verdict

Researchers entering robot manipulation or writing literature reviews will find the taxonomy and paper links invaluable. Pure practitioners looking for ready-to-run code should look elsewhere; this is a curated index, not a framework.