RoboLineage: Agent-Native Data Lifecycle Governance Across Robot Policy Iterations

2026-06-20

AI Summary

This work addresses the fragmented provenance of data across the robot policy iteration pipeline—spanning collection, curation, training, evaluation, and deployment—where evidence is scattered across heterogeneous tools and human memory, lacking unified governance. The paper introduces the first agent-centric lineage artifact model that explicitly represents each stage of policy iteration as typed, traceable artifacts, enabling end-to-end automated coordination. The framework integrates embodied replay parsing, data health monitoring, and cross-iteration state summarization, and is compatible with diverse robotic embodiments and training frameworks. Empirical results demonstrate that the approach significantly accelerates policy iteration in real-world tasks while preserving downstream performance and enhancing end-to-end auditability. The system has been released as an open-source, lightweight lifecycle management library.

Abstract

We present RoboLineage, an agent-native data lifecycle governance system for robot policy iteration. Modern robot policies improve through repeated data collection, review, retraining, evaluation, and release decisions, but the evidence connecting these steps is often scattered across local tools, scripts, and expert memory. RoboLineage makes this lifecycle explicit by representing rollouts, reviews, dataset decisions, training runs, policy metadata, evaluations, deployment recommendations, and next-collection plans as typed lineage artifacts. Agents interpret embodied rollout evidence, adapt accepted data to existing training stacks, maintain data health, and summarize cross-iteration state under explicit artifact boundaries. In real-robot manipulation workflows, RoboLineage makes routine policy iteration faster and more auditable while maintaining downstream policy performance. We open source RoboLineage as a lightweight lifecycle layer for different robot embodiments and training families. Project page: https://robolineage.github.io/

Problem

• data lifecycle governance
• robot policy iteration
• lineage tracking
• embodied AI
• auditability

Innovation

• agent-native
• data lineage
• robot policy iteration
• lifecycle governance
• embodied AI