Cheng Chi

Diffusion Policy Lead Author, UMI Developer, Sunday Robotics CTO

Profile

FieldDetails
Current PositionCo-founder & CTO, Sunday Robotics
PreviousColumbia University PhD, Stanford SNF
AdvisorShuran Song

Key Contributions

  • Diffusion Policy Lead Author: Applied diffusion to robot action generation (RSS 2023 Best Paper Finalist)
  • UMI (Universal Manipulation Interface): Universal manipulation data collection interface (RSS 2024 Best Systems Paper Finalist)
  • Iterative Residual Policy: Learning framework for repeatable tasks (RSS 2022 Best Paper Award)
  • Sunday Robotics Co-founder: Home robot startup CTO

Research Timeline

Columbia to Stanford to Sunday Robotics

Advised by Shuran Song

YearWorkImpact
2020NuroMapping & Localization Team
2021.01PhD BeginsColumbia CAIR Lab
2022Iterative Residual PolicyRSS 2022 Best Paper Award
2022DextAIRityRSS 2022 Best Systems Paper Finalist
2023Diffusion PolicyPioneering robot diffusion research (RSS 2023)
2024UMIRSS 2024 Best Systems Paper Finalist
2024Moved to StanfordTogether with Shuran Song (SNF)
2024.04Sunday Robotics Co-foundedWith Tony Zhao
2025.11Memo Robot LaunchHome robot unveiled

Major Publications

Diffusion Policy (RSS 2023, IJRR 2024)

“Diffusion Policy: Visuomotor Policy Learning via Action Diffusion”

Key contributions:

  • First application of diffusion to robot action generation
  • Handling multimodal action distributions
  • Average 46.9% performance improvement across 4 benchmarks

UMI (2024)

“Universal Manipulation Interface: In-The-Wild Robot Teaching Without In-The-Wild Robots”

Key contributions:

  • Data collection with human hands, without robots
  • Transfer to various robot platforms
  • In-the-wild data collection

Iterative Residual Policy (RSS 2022)

“Iterative Residual Policy for Goal-Conditioned Dynamic Manipulation of Deformable Objects”

Key contributions:

  • General learning framework for repeatable tasks
  • Can learn from inaccurate simulation data
  • RSS 2022 Best Paper Award, Best Student Paper Finalist

Key Ideas

Diffusion Policy (2023)

Core: Generate robot actions through denoising diffusion

p(a|o) = integral p(aK) product p(ak-1|ak, o) dak:K

Process:
1. Start from pure noise
2. Gradual denoising conditioned on observation (o)
3. Generate final action sequence

vs ACT:

AspectDiffusion PolicyACT
Generation MethodIterative denoisingSingle forward pass
MultimodalityNatural handlingStyle variable (z)
Training StabilityVery highHigh

UMI (2024)

Core: Collect data with human hands without robots, then transfer to robots

[Human hand demonstration] -> [UMI Interface] -> [Robot policy learning]

Advantages:
- Data collection without robot hardware
- Collection possible in real environments
- Transfer to various robots

Impact

Impact of Diffusion Policy

  • pi0: Adopted flow matching (a variant of diffusion)
  • Octo: Uses diffusion decoder
  • SmolVLA: Uses flow matching
  • LeRobot: Default supported model

Impact of UMI

  • Significantly reduced data collection costs
  • Influenced similar approaches like Sunday Robotics

Philosophy

Research Philosophy

“The key is combining good representations with good generative models”

Research Direction

  1. 2021-2023: Diffusion for robot learning
  2. 2024: Data collection interfaces (UMI)
  3. 2024-present: Sunday Robotics - Home robot commercialization

See Also