It turns out humans are really good at choosing parameters to fit a model. To use this approach, we need to choose a model structure, with some free parameters. We are going to apply the same command to the robot and to the model, and try to get the responses to match. We are going to manually search for good values for the model parameters to do this. If we can match actual data, we have a reasonable model.
Let's try to fit the velocities during command steps (steps8_1). We will choose an initial state (angle = 0 and angular velocity = 0), and apply the same commands we applied to the real robot. We will need to simulate this model.
Initially we are going to just focus on velocity data. The first model structure we are going to use is
1) if command > 0 torque = 0.0159*command*V/8.25 - 0.0784; else if command < 0 torque = 0.0157*command*V/8.25 + 0.0850; else % command = 0 torque = 0; 2) next_velocity = c0*velocity + c1*torque;The two unknown parameters are c0 and c1. Equation 1 is the model of the static relationship between commands and torque we came up with previously.
