{"version":"https://jsonfeed.org/version/1.1","title":"OpenRGD Engineering Journal","home_page_url":"https://openrgd.org/blog","feed_url":"https://openrgd.org/feed.json","description":"Source-backed engineering notes, architecture decisions, specification updates and research logs from OpenRGD.","language":"en","authors":[{"name":"OpenRGD","url":"https://openrgd.org"}],"items":[{"id":"https://openrgd.org/blog/so101-teleoperation-record-replay-evidence-boundary","url":"https://openrgd.org/blog/so101-teleoperation-record-replay-evidence-boundary","title":"SO-101 field note: teleoperation, one local episode, and what replay does not prove","summary":"A bounded OpenRGD SO-101 field note: operator-confirmed leader/follower teleoperation, one no-video 448-sample smoke episode, replay and the missing public evidence required for policy benchmarks.","content_text":"An operator-confirmed robot smoke test is worth reporting, provided it is not mistaken for a published benchmark. Our SO-101 leader/follower work established a native LeRobot record/replay path; the next job is to archive the evidence that makes it reproducible.","date_published":"2026-10-16T00:00:00Z","date_modified":"2026-10-16T00:00:00Z","tags":["Field Note","OpenRGD","robotics"]},{"id":"https://openrgd.org/blog/openrgd-0-2-0-why-robot-graph-definition-exists","url":"https://openrgd.org/blog/openrgd-0-2-0-why-robot-graph-definition-exists","title":"OpenRGD 0.2.0 — Why Robot Graph Definition exists","summary":"Why OpenRGD separates semantic embodiment from robot description, runtime execution and hardware authority — and what the 0.2.0 Draft actually defines.","content_text":"OpenRGD exists because describing a robot is not the same problem as grounding an intelligent system in that robot. Robot Graph Definition is our attempt to make embodiment explicit, machine-readable, source-aware and portable without quietly taking ownership of execution.","date_published":"2026-09-27T00:00:00Z","date_modified":"2026-09-27T00:00:00Z","tags":["Specification","OpenRGD","robotics"]},{"id":"https://openrgd.org/blog/robot-description-is-evidence-not-permission","url":"https://openrgd.org/blog/robot-description-is-evidence-not-permission","title":"A robot description is evidence, not permission","summary":"Why structural robot descriptions should inform embodied systems without silently authorizing control, safety assumptions or hardware execution.","content_text":"A robot file can tell us a great deal about a body. It cannot, by itself, tell an intelligent system that it is allowed to move that body. OpenRGD treats robot descriptions as evidence inputs, not execution authority.","date_published":"2026-09-27T00:00:00Z","date_modified":"2026-09-27T00:00:00Z","tags":["Architecture","OpenRGD","robotics"]},{"id":"https://openrgd.org/blog/urdf-to-openrgd-import-what-the-file-proves","url":"https://openrgd.org/blog/urdf-to-openrgd-import-what-the-file-proves","title":"From URDF to OpenRGD: importing what the file proves — and nothing else","summary":"A practical look at the OpenRGD URDF importer, its evidence boundary and the deliberate refusal to invent safety, identity, controllers or hardware bindings.","content_text":"The OpenRGD URDF importer has a deliberately narrow job: extract source-supported structural and physical evidence into a partial Foundation profile. Completeness is not the goal. Auditability is.","date_published":"2026-09-27T00:00:00Z","date_modified":"2026-09-27T00:00:00Z","tags":["Engineering","OpenRGD","robotics"]},{"id":"https://openrgd.org/blog/vla-actionintent-runtime-boundary","url":"https://openrgd.org/blog/vla-actionintent-runtime-boundary","title":"Why a VLA should not silently become the actuator authority","summary":"A candidate OpenRGD execution boundary for Vision-Language-Action systems: keep model output, embodiment grounding, safety validation and hardware authority distinct.","content_text":"Vision-Language-Action models make it increasingly natural for perception and language to produce robot actions. That does not mean model output should automatically become a motor command. OpenRGD treats the transition from cognition to execution as an explicit contract boundary.","date_published":"2026-09-27T00:00:00Z","date_modified":"2026-09-27T00:00:00Z","tags":["Physical AI","OpenRGD","robotics"]},{"id":"https://openrgd.org/blog/verified-does-not-mean-independently-tested","url":"https://openrgd.org/blog/verified-does-not-mean-independently-tested","title":"Verified does not mean independently tested","summary":"How OpenRGD separates source review, manufacturer claims, independent testing and provenance so a green verification state does not overclaim evidence.","content_text":"A robotics knowledge layer becomes dangerous when 'verified' quietly means 'true in every sense'. OpenRGD uses narrower evidence states: we can verify that a claim is supported by a reviewed source without pretending we independently measured the robot.","date_published":"2026-09-27T00:00:00Z","date_modified":"2026-09-27T00:00:00Z","tags":["Methodology","OpenRGD","robotics"]},{"id":"https://openrgd.org/blog/openrgd-runtime-boundary","url":"https://openrgd.org/blog/openrgd-runtime-boundary","title":"The OpenRGD Runtime Boundary","summary":"Why the canonical OpenRGD repository stops before physical execution, and how embodied runtimes and Body Adapters remain independently governed.","content_text":"One of the most important OpenRGD decisions is a negative one: the canonical repository does not own physical robot execution. The standard can describe, validate, compile and export context without becoming the process that energizes hardware.","date_published":"2026-09-27T00:00:00Z","date_modified":"2026-09-27T00:00:00Z","tags":["Architecture","OpenRGD","robotics"]}]}