Physical AI · Robotics · Engineering record

Engineering Physical AI Journal

A chronological record of moving from software systems into real robotics: the decisions, experiments, failures, measurements, and lessons worth preserving.

When a Wi-Fi Bug Starts Beeping

Directional Wi-Fi throughput, a low-battery plot twist, missing logs, and the cleaner experiment that came out of a robot debugging session.

Charging Became a Robotics Engineering Problem

Charging downtime, vendor power guidance, the external bench-PSU solution, and the questions it raised about power paths and autonomous docking.

Native ROS 2 Remote Development Instead of VNC

Why VNC is the wrong long-term abstraction and how an RTX 4090 Ubuntu workstation could become a native ROS 2 participant.

Preserve the Factory Baseline Before Modifying It

Hardware verification, ROS inventory, measured battery telemetry, SSH friction, and the strategy for a recoverable full-disk backup.

First Day with the Robot

Arrival, assembly friction, wiring questions, display issues, and the realization that robotics starts below ROS.

Preparing an RTX 4090 Robotics Workstation

Preparing Ubuntu, ROS 2, and an RTX 4090 workstation before depending on the robot-side Jetson as the only development machine.

Committing to the M3 Pro

The M3 Pro purchase decision, the risk I accepted, and the plan to verify, understand, back up, and eventually move beyond the vendor stack.

Choosing My First Physical AI Platform

The requirements, tradeoffs, security concerns, and Jetson decision behind choosing a mobile manipulator as a first robotics laboratory.

Why This Journal Exists

The two-year adaptation goal, ten-year horizon, and why failures and beginner questions belong in the permanent engineering record.