FedEMA: Federated Exponential Moving Averaging with Negative Entropy Regularizer in Autonomous Driving

TL;DR

FedEMA introduces a federated learning framework with exponential moving average and negative entropy regularization to improve autonomous driving models' generalization and continual adaptation in dynamic environments.

Contribution

The paper proposes FedEMA, a novel federated learning method that combines EMA-based model fusion and entropy regularization to enhance model robustness and prevent overfitting.

Findings

Achieves 7.12% higher mIoU on Cityscapes and Camvid datasets.

Effectively balances model generalization and adaptation.

Theoretical convergence of FedEMA is established.

Abstract

Street Scene Semantic Understanding (denoted as S3U) is a crucial but complex task for autonomous driving (AD) vehicles. Their inference models typically face poor generalization due to domain-shift. Federated Learning (FL) has emerged as a promising paradigm for enhancing the generalization of AD models through privacy-preserving distributed learning. However, these FL AD models face significant temporal catastrophic forgetting when deployed in dynamically evolving environments, where continuous adaptation causes abrupt erosion of historical knowledge. This paper proposes Federated Exponential Moving Average (FedEMA), a novel framework that addresses this challenge through two integral innovations: (I) Server-side model's historical fitting capability preservation via fusing current FL round's aggregation model and a proposed previous FL round's exponential moving average (EMA) model; (II) Vehicle-side negative entropy regularization to prevent FL models' possible overfitting to EMA-introduced temporal patterns. Above two strategies empower FedEMA a dual-objective optimization that balances model generalization and adaptability. In addition, we conduct theoretical convergence analysis for the proposed FedEMA. Extensive experiments both on Cityscapes dataset and Camvid dataset demonstrate FedEMA's superiority over existing approaches, showing 7.12% higher mean Intersection-over-Union (mIoU).