QERRA-v2 Classical is an open-source, fully explainable safety middleware layer designed to evaluate autonomous robot actions before execution.
The architecture uses a two-layer design:
- QERRA-HSR (Physical Safety Layer): A zero-ML, pure Python module that evaluates physical hazard proximity and human distress signals to trigger immediate fail-closed safety halts.
- SEMEV-12 (Moral Deliberation Engine): A hybrid semantic-pattern engine that evaluates prospective natural-language tasks against 12 human-centred ethical dimensions (such as coercion, gaslighting, or autonomy violations) with complete reasoning logs.
The system interfaces directly into ROS 2 pipelines via Action Servers and PyTrees Behavior Tree condition nodes using non-blocking background threading to ensure control loops never stall.
The simulation prototype, unit benchmarks, and 50 Hz telemetry test suites are complete. I’m looking for technical feedback, code reviews, and collaboration on simulation or physical hardware integration trials.
If you are interested in testing the node or exploring a joint trial, please reply here or contact directly at marunigno@gmail.com