Cookies help us deliver our services. By using our services, you agree to our use of cookies. Learn more

crowdAI is shutting down - please read our blog post for more information

NeurIPS 2018: AI for Prosthetics Challenge

Reinforcement learning with musculoskeletal models


Value of "target_vel" in observaton vector

Posted by Maksim_Kretov almost 2 years ago

Let’s say I set difficulty to 1. Observation vector is 160-dim (project=True) and should contain value target_vel somewhere. Where I can find it (which indexes exactly)? I can call get_state_desc() and see that target_vel is for example [1.25, 0, 0] but I don’t see any of these values in observation vector which is an input for the agent..

Posted by amirabdi  almost 2 years ago |  Quote

Check ProstheticsEnv.get_observation method in /home/amir/osim-rl/osim/env/osim.py You will see what the observation vector holds.

I kind of have the same problem with the action vector and I don’t know which muscle each of the 22 activations belong to.

Posted by Maksim_Kretov  almost 2 years ago |  Quote

Amirabdi, yes I saw this method, it exactly demonstrates my point – in observation vector there is no “target_vel” anywhere. Why? Is API going to change in Round 2?

Posted by amirabdi  almost 2 years ago |  Quote

@Maksim_Kretov I’m not sure. I am as well trying to figure out what the order of the muscle activations are in the action vector of length 22. Do you have any comments there?