It Takes Two Neurons to Ride a Bicycle
4 days ago
- #neural-network
- #robotics
- #bicycle-control
- A two-neuron network controls a bicycle by adjusting handlebar torque based on leaning angle.
- The first neuron outputs desired lean angle (γ), while the second neuron outputs handlebar torque.
- Inputs to the network include desired heading, current heading, lean angle, and their derivatives.
- The network uses a continuous-time model with thresholding functions on weighted sums.
- Indirect control of heading is achieved by regulating the lean angle via handlebar torque.
- The design avoids extreme leaning to prevent instability, mimicking real bicycle stability features.