Choosing My First Physical AI Platform
I want this journal to preserve the transition from software systems into robotics and Physical AI from the point where the decisions were still uncertain.
The first question was not "Which robot looks coolest?" It was what kind of platform would let me learn the most without spending months building basic mechanics before I could reach ROS, perception, navigation, manipulation, and AI.
My criteria became fairly specific:
- ROS 2 support and enough openness to inspect or replace the vendor stack
- a mobile base plus a real arm, not navigation-only hardware
- LiDAR and camera/depth sensing
- NVIDIA Jetson compute for future GPU robotics work
- enough preassembly to start engineering quickly
- hardware that can still be reused if I eventually outgrow the first robot
I compared chassis styles, including mecanum, and thought about the surfaces I actually care about: carpet, hard flooring, concrete, and small thresholds. I also compared Raspberry Pi-class systems with NVIDIA Jetson configurations and spent time deciding whether the stronger Jetson option was worth the cost.
The Jetson Orin NX 16 GB became an important part of the decision. Even if the first robot eventually becomes too limiting, the compute module can move into another project.
The configured M3 Pro price I was evaluating was around $2,350, which made the decision uncomfortable enough that I compared several alternatives and questioned whether the platform looked too much like an educational toy for the money.
Another concern was trust. The factory system is a vendor-prepared Linux/ROS image containing software and services I did not install. I do not assume that means it is unsafe, but a networked robot with cameras and sensors deserves scrutiny. My long-term plan became: verify the factory system, inventory it, back it up, and then gradually move toward an independently understood open Linux/ROS environment.
The decision point was ultimately this: I was not trying to buy the best finished robot. I was trying to buy a useful physical laboratory.
What I decided
The Yahboom ROSMASTER M3 Pro looked like a reasonable middle ground between a small educational platform and a much more expensive industrial or research robot.
The experiment would be to find out whether it could carry me far enough into real robotics before I needed the next platform.