Mistake in the implementation of current velocity array in the FillBatch method of the DataHandler class

[PFery4]

social_vrnn/src/data_utils/DataHandlerLSTM.py

Line 871 in 707cada

current_vel = np.array([trajectory.vel_vec[start_idx + tbp_step, 0], trajectory.vel_vec[

The definition of the current velocity array is incorrect. We can compare it with the way the current_pos array is implemented (which is the correct way), right above it

for prev_step in range(self.prev_horizon,-1,-1):
				current_pos = np.array([trajectory.pose_vec[start_idx + tbp_step - prev_step, 0], trajectory.pose_vec[
					                        start_idx + tbp_step - prev_step, 1]])
				current_vel = np.array([trajectory.vel_vec[start_idx + tbp_step, 0], trajectory.vel_vec[
					                        start_idx + tbp_step - prev_step - prev_step, 1]])

current_vel should be instead:

				current_vel = np.array([trajectory.vel_vec[start_idx + tbp_step - prev_step, 0], trajectory.vel_vec[
					                        start_idx + tbp_step - prev_step, 1]])

[bbrito]

@PFery4 thanks for finding it! I guess we did this mistake when we were cleaning up the code for release. Can you please make a pull request with the fix? I do not have edit privileges anymore so someone else has to merge it.

[PFery4]

I'm not super familiar with how pull requests work exactly, will give it a try in a few hours if I have the time.