QERRA-v2 Classical launches explainable ethical layer for ROS 2 humanoid robots
A 100% classical, auditable ethical evaluation engine based on 12 immutable human-centered vectors
QERRA-v2 Classical is an open-source ethical evaluation engine designed specifically for humanoid robots operating under ROS 2. Created by solo developer Marussa Metocharaki, it uses the SEMEV-12 framework, which defines 12 immutable human-centered ethical vectors. Unlike black-box AI safety systems, QERRA is 100% classical and fully auditable — every decision can be traced back to its ethical reasoning. The engine integrates as a Condition node in Behavior Trees, allowing robots to pause or redirect actions when conflicts arise. A ROS 2 bridge stub is included, which works standalone and becomes a full node when rclpy is available.
The current stable release is v1.8.1, with improved nuance handling for moral pressure and strong commitment cases. A live public API is hosted on Hugging Face, and a free demo key is available for testing. Metocharaki is now calling for serious early testers, especially from the ROS 2 and robotics community. Specific scenarios of interest include conflicting human instructions, medical or caregiving ethical dilemmas, and industrial high-stakes decisions. Testers are encouraged to submit prompts where the scoring or reasoning feels off, as honest feedback is critical. Interested parties can access the API, read the full call on GitHub, or email marunigno@gmail.com with the subject "QERRA Tester".
- 100% classical and explainable ethical layer using the SEMEV-12 framework with 12 immutable human-centered vectors
- Live public API on Hugging Face, stable v1.8.1 core, includes a ROS 2 bridge stub that becomes a full node with rclpy
- Seeking early testers for robotic scenarios involving conflicting instructions, medical/caregiving dilemmas, or industrial ethics
Why It Matters
As humanoid robots enter daily life, auditable ethical safety layers like QERRA are critical for trust and compliance.