Engineering notes from the robotics frontier.
Deep dives, post-mortems, and field reports from our team and the community.
How RoboAgent symbolically executes your launch files
Why pattern-matching .launch.py files breaks for any non-trivial workspace — and how we built a symbolic executor that respects every IfCondition, PathJoinSubstitution, and nested IncludeLaunchDescription.
The 9 QoS mismatches that silently break your robot
Best-effort vs reliable, late-joining subscribers, deadline misses. Plus the matrix the agent uses to flag them automatically.
Why every fix should pass simulation before it lands
Velocity without verification crashes robots. The case for closed-loop AI: propose, simulate, observe, iterate.
Cross-domain reasoning across micro-ROS
Your STM32 publishes BestEffort. Nav2 expects Reliable. The bug is invisible until the IDE understands both sides.
Designing the Robotics Knowledge Graph schema
How we ontologize ROS2 — what's a node, a topic, a frame, a controller, a hardware interface — and why structure beats embeddings.
Detecting TF drift before it kills your AMR
Why static_transform_publisher mistakes are still the #1 cause of warehouse robot pile-ups, and how the agent catches them in 4s.