ROSMASTER X3

Buying recommendations

Mapping a room with the 2D lidar is light work. Lidar SLAM (Cartographer or gmapping) and Nav2 navigation run comfortably on a Raspberry Pi 5. The Jetson boards earn their price on camera work. That means object detection with YOLO, pose tracking, 3D mapping from the depth camera with ORB-SLAM2 and Octomap, and small language models on the robot. Those jobs need an NVIDIA GPU with CUDA, which the Pi lacks. The software also differs. Yahboom's Pi 5 and Jetson Nano images use ROS 2 Foxy and ROS 1 Melodic, which both reached end of life in 2023. The Orin images use ROS 2 Humble, supported until 2027.

• Learning ROS, lidar mapping and navigation · Advanced with Raspberry Pi 5 ($1,019). The voice module costs only $30 more than Standard. Every mapping and navigation tutorial runs on the Pi. Heavier AI can run on a laptop that receives the camera stream over Wi-Fi.
• Camera AI and 3D mapping · Advanced with Jetson Orin Nano Super 8 GB ($1,369). The best value for running detection and depth mapping on the robot. It costs $50 more than the 4 GB board and has twice the GPU cores and memory. 4 GB fills up quickly with ROS and a vision model running together.
• Language models on the robot · Advanced with Jetson Orin NX Super 16 GB ($2,099). Worth it only when the robot must work without a network. Sending prompts to a laptop or cloud model costs less and gives better answers.
• Classrooms and demos · Superior with Raspberry Pi 5 ($1,129). The 7-inch screen adds $110 and shows the map on the robot itself.
• Owners of a spare Pi 5 or Jetson · Any kit without a main board. These kits cost $330 to $1,350 less than the same kit with a board.
• Versions to skip · Jetson Nano 4 GB and Ultimate. The Nano costs more than the Pi 5 for similar speed, and NVIDIA's software for it is stuck on Ubuntu 18.04. Ultimate swaps the A1 lidar for the longer-range S2L at about $310 extra with an Orin board. That only pays off in large spaces.