Committing to the M3 Pro
On August 20, 2026, I ordered the Yahboom ROSMASTER M3 Pro from RobotShop.
By that point the purchase was not an impulsive robotics gadget decision. I had already compared alternatives, questioned the price, thought about chassis behavior, looked at compute options, and considered whether I should instead build more of the system myself.
The M3 Pro won because it bundled enough of the robotics stack into one platform:
- mobile mecanum base
- robotic arm
- LiDAR
- camera/depth sensing
- ROS support
- NVIDIA Jetson Orin NX 16 GB
- enough preassembly to avoid spending the first stage only on mechanical construction
I also deliberately chose a platform with an arm. A base-only robot would have been useful for SLAM and navigation, but my longer-term interest is in machines that can both move through an environment and interact with it.
The risk I accepted
The platform was expensive enough that I knew I might later decide I had overpaid for something that felt educational rather than professional.
I accepted that risk for one reason: the value would not be measured only by the finished robot. The robot would be a laboratory for ROS 2, hardware, power, networking, perception, manipulation, edge AI, and eventually Physical AI.
The Jetson also reduced the downside. If I later replace the base, arm, or entire robot, the compute can potentially continue into another project.
Plan before delivery
I did not want to treat the factory software as a permanent black box. The initial plan was already:
- Verify every major component.
- Understand the factory ROS/Linux environment.
- Back up the original system before significant modification.
- Establish a proper remote development workflow.
- Gradually replace vendor dependence with software I understand and control.
The purchase was the point where the Physical AI plan stopped being only research and became a hardware commitment.