FedPAW: Federated Learning with Personalized Aggregation Weights for Urban Vehicle Speed Prediction

TL;DR

This work addresses privacy-preserving, personalized vehicle speed prediction in non-IID urban settings by introducing FedPAW, a federated learning framework that performs server-side personalized aggregation. FedPAW computes per-layer aggregation weights from the weighted parameter differences between local and global models, enabling tailored aggregated models to each client without extra client-side computation or communication. Using a CARLA-based CarlaVSP dataset and a Multi-Head Attention Augmented Seq2Seq LSTM predictor, FedPAW achieves state-of-the-art accuracy, notably a 0.8% MAE improvement over baselines at a 10-second horizon, while maintaining stability and reasonable overhead. The CarlaVSP dataset and FedPAW code are publicly available to facilitate future research in privacy-preserving, personalized autonomous-vehicle prediction.

Abstract

Vehicle speed prediction is crucial for intelligent transportation systems, promoting more reliable autonomous driving by accurately predicting future vehicle conditions. Due to variations in drivers' driving styles and vehicle types, speed predictions for different target vehicles may significantly differ. Existing methods may not realize personalized vehicle speed prediction while protecting drivers' data privacy. We propose a Federated learning framework with Personalized Aggregation Weights (FedPAW) to overcome these challenges. This method captures client-specific information by measuring the weighted mean squared error between the parameters of local models and global models. The server sends tailored aggregated models to clients instead of a single global model, without incurring additional computational and communication overhead for clients. To evaluate the effectiveness of FedPAW, we collected driving data in urban scenarios using the autonomous driving simulator CARLA, employing an LSTM-based Seq2Seq model with a multi-head attention mechanism to predict the future speed of target vehicles. The results demonstrate that our proposed FedPAW ranks lowest in prediction error within the time horizon of 10 seconds, with a 0.8% reduction in test MAE, compared to eleven representative benchmark baselines. The source code of FedPAW and dataset CarlaVSP are open-accessed at: https://github.com/heyuepeng/PFLlibVSP and https://pan.baidu.com/s/1qs8fxUvSPERV3C9i6pfUIw?pwd=tl3e.

Figures (10)

• Figure 1: Illustration of FedPAW on the server.
• Figure 2: Data collection scenarios. (a): Urban map Town 10 in Carla simulator. (b) and (c): RGB image and Semantic segmentation image captured by the camera sensor in Carla simulator.
• Figure 3: Vehicle speed distribution of 10 clients during driving (excluding parking).
• Figure 4: Schematic of V2V and V2I enabled traffic scenario.
• Figure 5: Multi-Head Attention Augmented Seq2Seq LSTM Model.