Release 1.1.0

New Features

1. Industrial Optimal Control Environments: We have added a collection of optimal control environments specifically designed for industrial applications. These environments enable you to train and evaluate reinforcement learning agents in realistic industrial scenarios.

2. sys_simulator Module: The new sys_simulator module is now available. This module allows you to simulate various systems and evaluate policies using different metrics, facilitating more comprehensive analysis and experimentation.

3. MPC Solver Integration: GOPS now integrates a Model Predictive Control (MPC) solver based on this package. This solver can efficiently determine the optimal policy for all 'model type' environments in GOPS. The optimal policy can serve as a reliable benchmark for policy trained by RL algorithms.

Enhancements

1. Unified Hyperparameters for MuJoCo Environments: We have carefully reviewed and unified the hyperparameters in the training examples for MuJoCo environments. This optimization ensures consistency and improved performance across different MuJoCo environments.

2. Improved DSAC Algorithm: The DSAC (Distributional Soft Actor-Critic) algorithm has been upgraded to a more stable and high-performance version.

3. Enhanced Efficiency and Stability of Parallel Trainer: We have made optimizations to the parallel trainer, resulting in improved efficiency and stability during training.