niufuxi / a-barrier-lyapunov-actor-critic-reinforcement-learning-approach-for-safe-and-stable-control

A-Barrier-Lyapunov-Actor-Critic-Reinforcement-Learning-Approach-for-Safe-and-Stable-Control

Repository containing the code for the paper "A Barrier-Lyapunov Actor-Critic Reinforcement Learning Approach for Safe and Stable Control", and this code is developed based on the code: https://github.com/yemam3/SAC-RCBF

This repository only contains the code for the algorithms Barrier-Lyapunov Actor-Critic (BLAC) and Barrier ActorCritic (BAC), for other algorithms, please refer to:

LAC: https://github.com/hithmh/Actor-critic-with-stability-guarantee

CPO, PPO-Lagrangian and TRPO-Lagrangian: https://github.com/openai/safety-starter-agents

An extension of this work is available at: https://arxiv.org/abs/2401.13148, accompanied by the code repository: https://github.com/LiqunZhao/Neural-ordinary-differential-equations-based-Lyapunov-Barrier-Actor-Critic-NLBAC. This repository features enhanced clarity through additional comments, a well-organized structure, and detailed instructions on easily applying the framework to your customized environments.

Installation Requirement

The experiments are run with Pytorch, and wandb (https://wandb.ai/site) is used to save the data and draw the graphs. To run the experiments, some packages are listed below with their versions (in my conda environment).

python: 3.6.13
pytorch: 1.10.2 
numpy: 1.17.5
wandb: 0.12.11
cvxpy: 1.1.20
gym: 0.15.7
gpytorch 1.6.0

Running the Experiments

The two environments are Unicycle and SimulatedCars. In Unicycle, a unicycle is required to arrive at the desired location, i.e., destination, while avoiding collisions with obstacles. SimulatedCars involves a chain of five cars following each other on a straight road. The goal is to control the velocity of the 4th car to keep a desired distance from the 3rd car while avoiding collisions with other cars.

Unicycle Env:

• BLAC algorithm: First navigate to the corresponding directory A-Barrier-Lyapunov-Actor-Critic-Reinforcement-Learning-Approach-for-Safe-and-Stable-Control/BLAC/Unicycle/, and then use the command: python main.py --env Unicycle --gamma_b 50 --max_episodes 50 --cuda --updates_per_step 2 --batch_size 128 --seed 0 --no_diff_qp --start_steps 1000

• BAC algorithm: First navigate to the corresponding directory A-Barrier-Lyapunov-Actor-Critic-Reinforcement-Learning-Approach-for-Safe-and-Stable-Control/BAC/Unicycle/, and then use the command: python main.py --env Unicycle --gamma_b 50 --max_episodes 50 --cuda --updates_per_step 2 --batch_size 128 --seed 0 --no_diff_qp --start_steps 1000

SimulatedCars Env:

• BLAC algorithm: First navigate to the corresponding directory A-Barrier-Lyapunov-Actor-Critic-Reinforcement-Learning-Approach-for-Safe-and-Stable-Control/BLAC/Simulated Car Following/, and then use the command: python main.py --env SimulatedCars --gamma_b 50 --max_episodes 50 --cuda --updates_per_step 2 --batch_size 128 --seed 0 --no_diff_qp --start_steps 200

• BAC algorithm: First navigate to the corresponding directory A-Barrier-Lyapunov-Actor-Critic-Reinforcement-Learning-Approach-for-Safe-and-Stable-Control/BAC/Simulated Car Following/, and then use the command: python main.py --env SimulatedCars --gamma_b 50 --max_episodes 50 --cuda --updates_per_step 2 --batch_size 128 --seed 0 --no_diff_qp --start_steps 200

For the meanings of the parameters, please refer to the main.py file.

Others

If you have some questions regarding the code or the paper, please do not hesitate to contact me by email. My email address is [email protected].