Robotics

Agnocast & Callback Isolated Executor: True Zero-Copy IPC and Middleware-Transparent Scheduling for ROS 2

New middleware and executor slash latency for autonomous vehicles, proven in real-world robotaxi deployments.

Deep Dive

The Autoware Foundation, the organization behind the open-source autonomous driving platform, has announced the full adoption of two new core technologies: Agnocast and the Callback Isolated Executor (CIE). Agnocast is a new, rclcpp-compatible middleware that enables true zero-copy inter-process communication (IPC) for all ROS 2 message types, including complex, unsized types like PointCloud2 that previously forced a fallback to slower serialization and copying. This breakthrough removes a major performance bottleneck in data-heavy autonomous systems. The Callback Isolated Executor is a drop-in replacement for the standard ROS 2 executor that establishes a persistent one-to-one mapping between each callback and a dedicated OS thread, eliminating the nested scheduling problem of traditional executors.

Proven in commercial robotaxi deployments, the technologies deliver dramatic real-world gains. Agnocast maintains consistent communication latency regardless of message size, a critical improvement for sensor data like LiDAR point clouds. Meanwhile, CIE enables per-callback OS scheduling control, which in live road tests improved the worst-case response times of critical control paths by approximately 5x compared to the default Linux scheduler. While they can be used independently, the tools are designed to cooperate; Agnocast requires a dedicated executor, a role filled by the CIE shipped with it. The packages are now available via the official ROS 2 Build Farm and a PPA, marking a significant step toward more deterministic, high-performance robotics software.

Key Points
  • Agnocast middleware enables true zero-copy IPC for ALL ROS 2 message types, including unsized PointCloud2, with no serialization overhead.
  • Callback Isolated Executor (CIE) maps each ROS callback to a dedicated OS thread, enabling direct OS-level scheduling control and predictability.
  • Live robotaxi tests showed a ~5x improvement in worst-case response times for critical control paths when using CIE.

Why It Matters

This dramatically improves the real-time performance and determinism of autonomous vehicles, making complex robotics systems faster and safer.