Observation on first results

Observation 1

If speed measurement is not well matched my the simulation, then the rest of the bicycle kinematic is poorly reproduced. weighting speed measurement in the cost function tends to work but compete with the other components of the objective function.

Observation 2

if no pure roll torque is introduced, the simulation uses huge (and unrealistic) amount of torso angle rates, this also lead to unrealistic Y acceleration patterns (even if this acceleration is still not used in the objective function).

Observation 3

-if an extra roll torque is introduced, this new control action dominates the control strategie, head X gyro matched well and the head acceleration (still not in the objective function) as well.

• What if the torso remains mostly vertical, controlled by small postural troque
• And most of the balance control is produced by legs, pushing on the pedals and producing a moment between the pelvis and the bicycle, this would produce a moment of force unrelated to torso motions.
• Question: can you produce a moment of forces between pelvis and the frame without relative motion between those two?
• Should we couple the extra roll torque to propelling torque? you can only produce the roll torque when push on the pedals....

The issue contains observations about speed weighting, torso motion, roll torque, and pedal coupling, but names no files, tests, or entry points. First clarify which control strategy is wanted and how it should be evaluated. Done would require an agreed scope and measurable criteria for the bicycle simulation and objective function.