Chapters
On this page
Papers and research
Start from a research question and follow links to the original paper, the authors' project page, and explanations on this site. The first entries cover work related to existing tutorials; topics such as locomotion, perception, and world models will be added over time.
Imitation learning and action generation
ACT · Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware
2023 · Bimanual manipulation / imitation learning / action chunking
Studies how low-cost bimanual hardware can learn fine-grained manipulation from demonstrations. ACT predicts a sequence of actions at once and is designed around compounding errors in long-horizon imitation learning.
- Original sources: paper · authors' project page
- Read on this site: ACT architecture and key mechanisms
- Try it: ACT bimanual training
Diffusion Policy · Visuomotor Policy Learning via Action Diffusion
2023 · Visuomotor policies / diffusion models / continuous actions
Models robot action generation as a conditional denoising process to represent multimodal action distributions, and combines visual conditioning with receding-horizon execution to generate action sequences.
- Original sources: paper · authors' project page
- Read on this site: Diffusion Policy action modeling
Vision-language-action models
OpenVLA · An Open-Source Vision-Language-Action Model
2024 · VLA / multi-robot data / fine-tuning
Connects a vision-language model to robot action prediction and releases the model and training code, making it easier to study cross-task learning and fine-tuning for new robot settings.
- Original sources: paper · authors' project page
- Read on this site: OpenVLA reproduction and key trade-offs
- Related resources: Embodied AI datasets · Open source projects
Compiled and sources checked: 2026-09-11.