**2. Related Work**

This section will introduce related works regarding sampling-based and velocity-based methods, DL and DRL. Sampling-based and velocity-based methods are classical navigation methods that are not suitable for dealing with environmental changes. Learning-based methods such as DL and DRL can improve the learning ability of mobile robots to quickly adapt to new environments. DL requires considerable amounts of labeled data, which is difficult to be collected, while DRL does not need prior knowledge.
