hello, when testing on TurtleBot dataset, what parameters should I choose? '--filt

Parameters configuration about TurtleBot dataset about learned-map-prior HOT 1 OPEN

klabcmu commented on August 28, 2024

Parameters configuration about TurtleBot dataset

from learned-map-prior.

Comments (1)

karnikram commented on August 28, 2024

Hi,

'--filter-history-length' and '--data-kernel-len' should be set to 20, right?

Yes.

And how about the '--filter-update-rate'? Is it 60Hz or 100Hz?

It is 100 Hz.

What's more, can you describe how to transform wheel odometry to right heading in detail?

To obtain heading from wheel odometry data we use a standard Madgwick filter such as this one. Once you have the heading, we apply a motion model as described in this code snippet:

def wheel_odom_motion_model(x_t0, odom, odometry_cov):
    current_heading_angle = x_t0[2]
    step_size = np.linalg.norm(odom[:2])
    step = np.array([step_size*np.cos(current_heading_angle), step_size*np.sin(current_heading_angle)])
    noise = sample(odometry_cov)
    step = step + np.random.normal(0,0.01)
    
    heading_delta = odom[2] + np.random.normal(0,0.01)
    
    x_t1 = np.zeros_like(x_t0)
    x_t1[:2] = x_t0[:2] + step

    x_t1[2] = current_heading_angle + heading_delta
    return x_t1