The open project AxonOS has been presented, developing software infrastructure for systems with a neurocomputer interface, designed to create a deterministic execution layer between neural hardware and AI. The AxonOS kernel is being developed in Rust in `no_std` mode, emphasizing predictable real-time execution, a formalized ABI for system calls, a capability-based access model, without blocking IPC, and runtime control. The source code is open under MIT or Apache 2.0 licenses.
The project includes a kernel/runtime, BCI integration, architectural specifications, a permissions/consent-based access model, and real-time analysis tools. Currently, AxonOS is being developed for ARM Cortex-M processors and is in the process of verifying functionality on hardware and independently testing real-time characteristics. The AxonOS architecture considers neurocomputer interfaces as a multi-layered system: neurocomputer device → HAL → deterministic runtime → IPC → permissions/consent access model → application and AI layer.
The core idea of the project is to create a controlled application execution layer functioning between the neurocomputer device and AI, where requirements for determinism, isolation, access to neural data, and response time are part of the system architecture rather than a function of the application. A separate direction of the project is analyzing the worst-case response time (WCRT) for tasks with fixed priorities. For this purpose, an independent tool called 'dy-wcet' is being developed — a 'no_std' Rust crate implemented without floating-point operations.
Source: opennet.ru
