MEFCL：Multiple Environment Federated Continual Learning
The Multi-Environment Federated Continual Learning (MEFCL) algorithm explores how intelligent agents can quickly adapt to new environments while not forgetting old environment knowledge, and promotes knowledge fusion across multiple environments.
Model Architecture Diagram：
Description： The MEFCL algorithm explores how an intelligent agent can adapt quickly to new environments without forgetting old environment knowledge and promotes knowledge fusion among multiple environments. During the federated training phase, the MEFCL algorithm uses a regularization term to constrain the updates of important parameters of the model, improving the model’s memory while using a dual-model training method to promote the fusion of the agent’s old environment knowledge with the new environment. When the terminal environment changes, the MEFCL algorithm quickly adapts to the new environment by replacing the environment information layer.