Feature learning dataset

该数据集是通过随机模拟来训练特征和逆环境动力学模型的。针对倒立摆环境，该数据集不仅结合了随机模拟，还融入了专家策略生成的模拟，以保持摆杆的平衡。由于在随机模拟中摆杆容易迅速下落，因此该数据集特别为倒立摆环境进行了定制。它适用于高维连续环境，任务是训练特征编码器和逆环境动力学模型。

This dataset is constructed via random simulation for training feature encoders and inverse environmental dynamics models. For the inverted pendulum environment, this dataset not only incorporates random simulation but also integrates simulations generated by expert policies to maintain the balance of the pendulum. Since the pendulum tends to fall rapidly in random simulations, this dataset is specifically tailored for the inverted pendulum environment. It is applicable to high-dimensional continuous environments, with the core task of training feature encoders and inverse environmental dynamics models.