bug in climbing and descending stairs

[bottlele]

Hello, in this project, using pre-trained weights works fine for climbing stairs, but when descending stairs, the model tends to step on the edge at the second-to-last step, leading to a fall. How can this issue be resolved? Thank you for your reply.

down.mp4

up.mp4

[yinkangning0124]

Hi, have you face the problem that the robot can not climb a single step, just stepping inplace?

[yinkangning0124]

climb_localmini.mp4

[bottlele]

My previous issue was about passing directly through the stairs instead of walking on them.

[yinkangning0124]

My previous issue was about passing directly through the stairs instead of walking on them.

This issue I encountered before, may be you can check your mujoco version, the version should be 2.2.x

[yinkangning0124]

Maybe we can communicate more deeply about this project. Feel free to contact me through yinkn2022@shanghaitech.edu.cn

[bottlele]

I see you are in Shanghai, are you Chinese?

[yinkangning0124]

I see you are in Shanghai, are you Chinese?

yeah, a master student at shanghaitech university

[bottlele]

I am currently in Changning District, Shanghai too

[WangXingfang]

My previous issue was about passing directly through the stairs instead of walking on them.

This issue I encountered before, may be you can check your mujoco version, the version should be 2.2.x

请问您程序里mujoco和dm-control的具体版本是多少呢？我这边运行的话，楼梯显示不出来

[WangXingfang]

Hello, in this project, using pre-trained weights works fine for climbing stairs, but when descending stairs, the model tends to step on the edge at the second-to-last step, leading to a fall. How can this issue be resolved? Thank you for your reply.

请问您程序里mujoco和dm-control的具体版本是多少呢？我这边运行的话，楼梯显示不出来，更换mujoco和dm-control的版本后有时候能显示出楼梯来，之后运行又显示不出来了

[skr3178]

@bottlele
what command do you use to run up the stair using pretrained weights?
also the command for running down the stairs shows different scenarios- walking sideways etc- on running it each time show different scene

PYTHONPATH=.:$PYTHONPATH python scripts/debug_stepper.py --path /xxxxx/LearningHumanoidWalking/trained/jvrc_stepper/critic.pt

Traceback (most recent call last):
File "scripts/debug_stepper.py", line 159, in
main()
File "scripts/debug_stepper.py", line 155, in main
run(env, policy, args)
File "scripts/debug_stepper.py", line 92, in run
action = policy.forward(torch.Tensor(observation), deterministic=True).detach().numpy()
TypeError: forward() got an unexpected keyword argument 'deterministic'