Large-Data-Driven Manipulation
Data engines, control, and learning for manufacturing automation
How data, control, and deployment reshape manufacturing manipulation. — 3 Parts, 12 Chapters
First published: 2026-06-18 | Last updated: 2026-07-22
Manufacturing First
Frames manipulation through factory cells, quality gates, failure logs, and repeatable work rather than demos alone.
Data Flywheels
Explains how human demonstrations, robot fleets, simulation, force/touch, and QA traces become one learning loop.
Research and Startups
Maps the Tedrake, Finn, Abbeel, and Levine lineages alongside US startup strategies.
Part I: The Data Problem
Data Scale — Conditions for Manufacturing Coverage
Defines high-mix, high-variance, contact-rich work as a data problem.
→ 02End Effectors — Designing Data Difficulty
Compares how end-effectors reshape data collection and learning difficulty.
→ 03Data Flywheels — Accumulating Learning Assets
Defines the task data and evaluation harness manufacturers must own.
→Part II: Control and Learning
Contact Models — Foundations for Control and Transfer
Explains how model-based control and contact modeling support large-data strategies.
→ 05Imitation and Reinforcement Learning — From Data to Policy
Connects robot data collection, imitation learning, reinforcement learning, and offline learning.
→ 06Policy Architectures — Balancing Generality and Execution
Maps policy architectures that include VLAs without reducing the field to VLAs.
→ 07Human Demonstrations — Translating Work into Robot Action
Compares strategies for turning human hand and work data into robot-executable data.
→ 08Contact Learning — Online Improvement Through Rich Sensing
Explains why vision-only scaling needs tactile and force-rich data.
→Part III: Deployment Strategy
Foundation Models — Data Strategies for Generalist Policies
Compares the data strategies of US startups building general physical AI models.
→ 10Production Data — Deployment as an Improvement Loop
Analyzes how deployed production data becomes model improvement and deployment moat.
→ 11Hardware Co-Design — Building for Learnability
Compares data gloves, custom hands, five-finger hands, and touch/force integration.
→ 12Manufacturing Data Sovereignty — What Operators Must Own
Defines buy/build criteria around replay sets, update governance, and data rights.
→