Unlock Reliable Skill Inference for Quadruped Adaptive Behavior by Skill Graph

The code repository contains relevant configuration requirements, fundamental skills training, RSG construction, inference and composition code. This repository is based off of Nikita Rudin's legged_gym and AMP repo, and enables us to train policies using Isaac Gym.

1.1 CODE STRUCTURE

1. Each env is defined by an env file (legged_gym/envs/base/legged_robot.py) and a config file (such as legged_gym/envs/a1/a1_amp_forward_walking_config.py). The config file contains two classes: one conatianing all the environment parameters (LeggedRobotCfg) and one for the training parameters (LeggedRobotCfgPPo).
2. Both env and config classes use inheritance.
3. Each non-zero reward scale specified in cfg will add a function with a corresponding name to the list of elements which will be summed to get the total reward. The AMP reward parameters are defined in LeggedRobotCfgPPO, as well as the path to the reference data.
4. Tasks must be registered using task_registry.register(name, EnvClass, EnvConfig, TrainConfig). This is done in legged_gym/envs/__init__.py.
5. Skill construction code can be found in the rsg_construction folder.

Usage

1. Train a single fundamental skill

1.1 Installation

1. Create a new python virtual env with python 3.6, 3.7 or 3.8 (3.8 recommended). i.e. with conda:
   • conda create -n sg python==3.8
   • conda activate sg
2. Install pytorch 1.10 with cuda-11.3:
   • pip3 install torch==1.10.0+cu113 torchvision==0.11.1+cu113 tensorboard==2.8.0 pybullet==3.2.1 opencv-python==4.5.5.64 torchaudio==0.10.0+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html
3. Install Isaac Gym
   • Download and install Isaac Gym Preview 3 (Preview 2 will not work!) from https://developer.nvidia.com/isaac-gym
   • cd isaacgym/python && pip install -e .
   • Try running an example cd examples && python 1080_balls_of_solitude.py
   • For troubleshooting check docs (isaacgym/docs/index.html)
4. Install rsl_rl (PPO implementation)
   • Clone this repository
   • cd AMP_for_hardware/rsl_rl && pip install -e .
5. Install legged_gym
   • cd ../ && pip install -e .

1.2 Not using AMP:

CUDA_VISIBLE_DEVICES=0 python legged_gym/scripts/train.py --task=a1_amp_forward_walking --actor_critic_class=ActorCritic --terrain_id=16 --num_envs=3000 --max_iterations=5000 --isObservationEstimation --isEnvBaseline --headless

CUDA_VISIBLE_DEVICES: Specify the GPU device on which the program is running.

--task: different task.

--actor_critic_class: utilizing Actor-Critic framework (PPO).

--terrain_id: different terrain.

--num_envs: the number of environments in parallel.

--max_iterations: the number of PPO algorithm iterations.

--isObservationEstimation: context-aided estimator network (CENet) architecture.

--isEnvBaseline: goal-conditional policy.

--headless: Does not display the graphical interface.

1.3 Using AMP:

The realistic demonstrations data for AMP are available google drive.

CUDA_VISIBLE_DEVICES=0 python legged_gym/scripts/train.py --task=a1_amp_forward_walking --actor_critic_class=ActorCritic --skills_descriptor_id=5 --terrain_id=0 --headless

--skills_descriptor_id: Differential weighting of intrinsic and extrinsic rewards.

2. RSG Construction

2.1 Installation

Setup python virtualenv and install packages as following:

cd rsg_construction/
python -m venv ./venv
source ./venv/bin/activate
pip install -r requirements.txt

2.2 Training and evaluation

The trained skills and task descriptions are available google drive.

Train and evaluate the built RSG:

cd rsg_construction/
python train.py

and

cd rsg_construction/
python test.py

3. Skill inference and composition

3.1 Skill inference and execution

python legged_gym/scripts/train.py --task=a1_amp_ct_b_sequential_2 --terrain_id=18 --num_envs=1 --max_iterations=200 --isObservationEstimation --case_id=2

3.2 Skill composition

python legged_gym/scripts/train.py --task=a1_amp_ct_b_sequential_3 --terrain_id=18 --num_envs=1 --max_iterations=200 --isObservationEstimation --case_id=0

The skill inference is implemented in class BOOnPolicyRunnerSequentialCase1 in file rsl_rl/rsl_rl/runners/sg_on_policy_runner.py.

The skill composition is implemented in class NewCompositeActor in file rsl_rl/rsl_rl/modules/composite_actor_bo.py.

The BO method is implemented in class CompositeBO in file rsl_rl/rsl_rl/algorithms/ppo.py.