Zenoh-Flow drift improving an Autonomous Driving platform
summary
Zenoh-Flow ports the Pylot autonomous driving pipeline onto Zenoh by replacing the airdosh layer and running a declarative YAML graph that the daemon deploys location transparently. It records timestamps and uses logical clocks plus unique IDs for total ordering and reproducible inputs, targets low latency with zero copy and reports 1.6M msgs/s with 8 bit payloads on benchmarks. A Carla demo drives a vehicle, shows simulation 100s vs system ~20s on v0.3, and offers Rust and Python bindings with C++ and ROS2 planned.
AI-written summary · about
chapters (7)
- 0:00Start
- 6:00Unified abstraction
- 6:44Multiple bindings
- 6:56Declarative
- 7:54Performant
- 8:39Automated deployment
- 9:04Features
related talks
01h 03mZenoh: The Edge Computing Data Fabric
36mEclipse Zenoh
26mFreeFlow: Software-based Virtual RDMA Networking for Containerized Clouds
28mScaling DDS and MQTT with Eclipse Zenoh
01h 02mHydroflow: A Compiler Target for Fast, Correct Distributed Programs
38mSelf-managed and automatically reconfigurable stream processing
30mData Platform Infrastructure at Rivian: A Deep Dive
36mScaling robots -