Averiom

Averiom is an edge-native safety system for physical AI. It learns constraint boundaries from expert operator micro-corrections, predicts boundary violations 150-300 milliseconds ahead, and enforces safe envelopes at the edge in under 5 milliseconds.

The system addresses limitations of rule-based safety systems that generate false positives and encourage workarounds, as well as the risks of manual operations where skilled operators may fail under fatigue or pressure. It infers the Negative Manifold, a 3D region of failure space, rather than prescribing rigid paths or motions. A prediction kernel applies Kalman filtering to forecast trajectories while an inhibition engine delivers tiered responses ranging from haptic warnings to viscous damping and hard locks via solenoid. All processing occurs locally with high-frequency fusion of vision, IMU, and force sensor data at 1 kHz to detect micro-hesitations and tremors. Every intervention produces cryptographically signed, tamper-evident records stored in portable .avm profiles, forming an evidence chain for forensic audit, liability protection, and transferable skill documentation.

It targets autonomous and semi-autonomous mining equipment in Queensland for hydraulic arm control on haul trucks and excavators. Additional focus areas include precision welding and carbon fibre layup for composites manufacturing, including AUKUS submarine production in Australia. The platform remains hardware-agnostic with zero cloud dependency. Future applications are noted in medical training, surgical workflows, consumer tools, and sports coaching. Averiom has reached Technology Readiness Level 4 with a proof-of-concept demonstration showing 182.5 ms end-to-end latency, zero false negatives in testing, and 100 percent detection rate. It holds a US provisional patent application filed 18 January 2026 and has peer-reviewed publications including an accepted paper for Springer CCIS and presentation at HCI International 2026. The project collaborates with the Australian Government Industry Growth Programme, Australian Manufacturing Capability Network, and Australian Manufacturing Growth Centre.