FedCKA: Representation-Guided Layer Personalization for Federated 3D Perception Across Driving Domains

2026-10-01

AI Summary

This study addresses the challenges of cross-driving-domain data heterogeneity in federated learning and the limited adaptability of conventional personalization methods caused by fixed layer partitioning. To overcome these limitations, this work proposes an adaptive layer-wise personalization strategy based on Centered Kernel Alignment (CKA). By transcending predefined layer partitions, the proposed method leverages CKA to dynamically measure layer-wise representation similarity and generate aggregation masks, thereby enabling an adaptive trade-off between personalized and globalized learning through selective parameter aggregation. Evaluated on the nuScenes multi-domain benchmark, the approach achieves a 7-percentage-point improvement in average NDS over the strongest baseline, demonstrating its superior robust perception capabilities in complex environments.

Abstract

Robust perception in intelligent vehicles demands 3D object detectors that remain dependable under domain shifts, such as changes in time of day, location, or weather. However, due to costly annotation and rare shifts, some environments lack sufficient data to train a standalone detector. Federated learning offers a privacy-preserving framework for collaborative model training, enabling clients to benefit from shared learning across diverse environments. Yet, this framework traditionally relies on a single global consensus model, which struggles to perform across heterogeneous local data distributions. Local conditions are better captured by adapting a subset of the model, but many personalization approaches rely on predefined layer partitions or fixed personalization ratios, thereby limiting adaptation to client-specific divergence. To reduce this rigidity, we propose FedCKA, a Centered Kernel Alignment (CKA)-based strategy that dynamically handles the personalization-globalization trade-off. Specifically, FedCKA computes layer-wise feature similarities between local client models and the global consensus model during training. By converting layer-wise similarity scores into client-specific aggregation masks, FedCKA selectively shares representation-consistent layers. Evaluation on a unified multi-domain benchmark based on nuScenes shows that FedCKA outperforms established federated baselines, including FedBN, FedRep, and FedSelect, improving average NDS by 7 percentage points over the strongest baseline. The findings offer both a comparative benchmark and a promising direction for robust federated 3D perception across shifts in location, weather, and illumination. Code is available at https://github.com/j-verhoog/FedCKA.

Problem

• Federated Learning
• 3D Object Detection
• Domain Shift
• Model Personalization
• Autonomous Driving

Innovation

• Federated Learning
• Centered Kernel Alignment
• 3D Perception
• Layer Personalization
• Domain Shift