Make Your Own Humanoid Robot

They may have a long way to go before they live up to the lofty promises of sci-fi movies, but even so, humanoid robots are rapidly stepping out of the laboratory and into the real world. However, if you want to get your hands on one, you will have to brace yourself for some serious sticker shock — they don’t come cheap. Fortunately, this is beginning to change. Advances in AI, in particular, are making these machines more accessible.

For instance, consider the Samuroid, a humanoid robot developed by an engineer who goes by the handle “alisa.wu.” It is built around inexpensive and accessible hardware, and it runs a wide range of AI algorithms to enable advanced capabilities. Samuroid is completely open source, so you can build your own without breaking the bank.

Samuroid is a 22-degree-of-freedom bipedal humanoid designed for experimentation with embodied AI and real-time locomotion. The chassis is constructed from high-strength aluminum alloy, giving the 2.3-kilogram robot enough rigidity to handle dynamic balancing while remaining relatively lightweight. Its joints are powered by high-voltage serial bus servos rated at 30 kgf·cm of torque at 12 volts, providing enough strength for expressive gestures, stable walking, and even energetic preset actions like kicking and dancing.

The robot’s brain is a Raspberry Pi 4B with 4GB of RAM, running Ubuntu 18.04 and fully integrated with ROS Melodic. This combination gives developers access to a mature robotics ecosystem, complete with tools for motion planning, sensor integration, and real-time control. Samuroid’s locomotion engine relies on inverse kinematics paired with the Linear Inverted Pendulum Model, helping maintain a stable center of mass during walking and other transitions.

A dedicated driver board manages power distribution, servo communication, and audio output, creating an architecture that separates high-level reasoning from low-level motor control. On the sensing side, Samuroid is equipped with a 1080p wide-angle camera, a 6-axis IMU based on the MPU6050, and a high-precision microphone with acoustic echo cancellation. Using OpenCV, the robot can perform facial recognition, color tracking, and even automatic ball targeting.

By connecting to APIs such as DeepSeek, the robot can interpret natural language commands and translate them into physical actions. A phrase like “show me something fun” can be semantically parsed into an appropriate behavior — say, triggering a dance routine via a ROS action server — before delivering spoken feedback through its onboard speaker.

With fully broken-out GPIO pins for expansion and compatibility with dozens of modular sensors, Samuroid is not just a hobby project, but a serious research and hacking platform. If you’d like to learn more, take a look at the full project write-up.