Design PrinciplesΒΆ

When designing Miha we have considered the following principles:

  1. Affordable: Traditionally, complex humanoid robots are expensive and difficult to afford. Costs tend to grow exponentially with the size of the robot and, while the costs of electronics follows the same downward trajectory that applies to other consumer electronic products, the mechanical parts do no not exhibit such a trend. We therefore have considered a design that offers enough volume to permit high performance computing at the edge while still minimizing the requirements for the actuators.

  2. Complex but not complicated: Toy robots are fun but useless when designing complex ML models or robotics frameworks. To provide utility, the robots need to have a degree of complexity that will warrant innovative ML models and robotics frameworks. We believe 22 DoF is a minimum that reflects the need for studying bipedal locomotion and interaction. A good array of sensors (position, effort, vision, sound, etc.) and processing abilities need to complement its high-performing actuators.

  3. Easily serviceable and expandable: Humanoid robots have a very tough life. Because a lot of the research is still in infancy, accidents happen and robots break quite often. Many of the platforms available on the market are not user serviceable which means significant downtime in research until the robots are returned from service. Miha is designed to be be easily serviced by any user with some minimum technical skills: parts are easily available and standard (Raspberry Pi, Dynamixel Servos) and frames are 3D printable or available as spares. Due to the modular nature of the frames, it is also very easy for users to design and 3D print custom parts that would provide the required functionally for a project at hand.

  4. Open and based on standardized framework: We aim to release as much as possible from the design of the robot as open-source (hardware and software) and we will encourage the community to contribute and expand this base. We are integrating standard frameworks (ROS, TFLite, etc.) with the robot.