In most software projects, a failing test means a bug. In robotics, it can mean a broken robot.
The pib humanoid robot is a community-driven, open-source humanoid platform used in schools, universities, vocational training, and maker spaces. With 28 degrees of freedom, AI capabilities, camera-based perception, and speech output, it allows students, researchers, and makers to build, program, and experiment with their own humanoid robot, often using parts they printed themselves.
But how do you safely test software that controls motors, balance, and physical movement?
The user interface software Cerebra communicates with the robot through Robot Operating System (ROS), orchestrating perception, behavior, and motion. Traditional automated testing alone is not enough: some failures only become visible when a robot moves, speaks, or falls.
This talk shows how we test such a system without constantly risking expensive hardware. A custom hardware-in-the-loop test stand allows us to run tests directly against the robot’s electronics and motors before deploying them to a full robot. Only when those tests pass do we move to real-world validation.
Along the way, we also test hardware prototypes and new locomotion concepts, like a humanoid that balances and moves like a Segway using PID-controlled legs.
Expect practical insights from the intersection of robotics, hardware testing, and software quality, where every test might move a robot, and every failure might send it crashing. 🤖