Heading task results not what I expected..

[ameliacode]

After training heading task with custom low-level policy, character seems to move but not like what it seemed in the paper.

I did trained my low level policy with possible locomotion direction. locomotion task showed good result, so I don't think re-training or adding data won't be a solution

What could be the problem?

[xbpeng]

what does the behavior of the low-level policy look like? Does sampling random latents for the low-level policy lead to more natural behaviors?

[ameliacode]

well I did solved this by highering saving frequency. (better result than above,,) AMP seems doing its task, however, it doesn't seem to "walk" while heading randomly. ASE performed it worse.. Other tasks worked well in ASE, but Heading tasked didn;t do the work

I'll update as soon as my experiment is done