Skip to content

M4: Reusable Components, Security Hardening, and Performance Optimization #766

Open

Description

@ytakano

Refactor the sensing and localization components into reusable, platform-independent libraries that can run on Awkernel and other environments. Improve security, robustness, and performance while preserving validated behavior and reusing the evaluation infrastructure established in M1–M3.

Scope:

  • Build on the existing realtime_core libraries and consolidate reusable functionality without duplicating implementations
  • Separate algorithmic cores from platform-specific scheduling, communication, timing, logging, and synchronization adapters
  • Define APIs, ownership rules, initialization and reset behavior, and memory and resource contracts
  • Demonstrate reuse on Awkernel and a Linux-hosted test application
  • Harden input validation, resource limits, and unsafe or FFI boundaries
  • Add fuzz testing, property-based testing, and regression cases for malformed and extreme inputs
  • Optimize measured bottlenecks such as copying, allocation, data layout, and synchronization
  • Establish shared regression tests and CI coverage for supported environments

Completion criteria:

  • Shared cores can be built and used without Awkernel-specific dependencies, with documented platform requirements and feature configurations.
  • The same core implementations are exercised through Awkernel and Linux adapters using shared test datasets.
  • Behavior-preserving extraction passes the applicable M1–M3 comparison and regression tests within predefined tolerances.
  • Input-validation rules, resource bounds, and unsafe or FFI assumptions are documented and tested against a defined threat model.
  • Security findings are resolved or explicitly documented with their impact and mitigation.
  • Performance changes are evaluated against a fixed baseline, reporting accuracy, tail latency, maximum observed execution time, allocation and copy counts, and peak memory usage where applicable.
  • Optimization targets and permitted regressions are defined before evaluation; trade-offs are explicitly reported.
  • APIs, integration examples, build configurations, and reproduction steps are documented.

Implementation approach:

Separate behavior-preserving extraction from optimization and intentional behavior changes. Architecture work may begin before M3 is complete, but final acceptance must use the relevant M1–M3 validation assets. Essential input safety and known vulnerability fixes must not be deferred until M4.

Out of scope:

New perception or planning functionality, mandatory ROS 2 runtime dependencies in shared cores, kernel-level fault isolation, formal security certification, and formal worst-case execution-time guarantees.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions