I designed a cube that balances itself on a corner
A home-built “balancing cube” that can stand on one edge or corner using internal reaction wheels has captivated readers both for its surreal behavior and its precision metalwork. Commenters explore how the control approach relates to satellite attitude systems and gyroscopic transports, debate safety and cost trade-offs (the current build is around €2,500 in parts), and swap ideas for cheaper variants using 3D‑printed flywheels, drone motors, or fluid-based actuators. The project also prompts broader reflections on open-sourcing academic work, learning electronics and control theory as a software engineer, and the potential for turning such mechanisms into art objects or commercial desk toys.
Overall reaction
- Many commenters find the object mesmerizing and beautiful, likening it to “magic” or fantasy artifacts.
- Strong desire to buy or mass‑produce a cheaper version; some want museum‑scale or permanent public sculptures.
- Skepticism appears around practicality, cost, and safety (high‑speed flywheels near fingers, potential for violent failures).
Cost, build & safety
- BOM reported around €2500; motors and controllers alone are several hundred euros, plus costly CNC machining.
- Suggestions to reduce cost: 3D‑printed flywheels with added weights, drone/quadrotor motors and ESCs, waterjet‑cut steel instead of milled stainless, laser‑cut acrylic panels.
- Cheap Cubli‑like kits and modules exist on Chinese marketplaces, from ~€50–500, but quality and behavior are unclear.
- Safety ideas: plexiglass or steel shields, fully enclosed cubes, attention to vibration at thousands of RPM; concern that 3D‑printed wheels may be weaker than steel.
Mechanism & physics
- Core principle: reaction wheels providing torque to change orientation; closely related to spacecraft reaction wheels.
- Clarification that simply shifting center of mass (e.g., fluid sloshing) is different; pumped‑fluid systems can mimic reaction wheels but are less practical for constant rotation.
- Distinction between reaction wheels and control‑moment gyroscopes (used in gyro‑monorails).
- Discussion of wheel saturation and need to slow or reverse wheels to maintain control.
Control & software aspects
- Balancing is a classic control problem (IMU + feedback controller). Many mention PID tuning, sign errors, and the inevitability of trial‑and‑error even with solid theory.
- Fusion 360 is used to get inertias and masses for the control model, but gains were still hand‑tuned.
- Some propose neural‑network controllers as an alternative to manual tuning.
Related devices & applications
- Direct connections to satellites, ships’ stabilizers, gyro‑monorails, self‑balancing robots, and art pieces like a self‑balancing couch.
- Plans and wishes for jump‑up and “walking” behavior using mechanical brakes; references to prior Cubli work and M‑Blocks.
- Speculative applications: vehicle and building stabilization, desktop toys, climbing a large cube, solar‑powered or inductively charged versions.
Learning, accessibility & meta
- Many software engineers express envy; others counter that Arduino/ESP projects and online kits make this approachable.
- Some debate whether this counts as “design” vs “invention,” but most agree the mechanical and control design work is substantial.
- A long tangent explores broader AI/automation impacts and “swarms” of bots, with pushback that machines still struggle outside constrained environments.