FedSceneX: Time-to-Target Orchestration for Same-Scene Multimodal Federated Edge Learning

2026-08-07

AI Summary

Traditional federated learning evaluates performance primarily in terms of communication rounds, overlooking the heterogeneous training and communication costs across modalities such as cameras, LiDAR, and radar, which leads to biased assessment and inefficient resource utilization. This work proposes FedSceneX, a time-aware scheduling framework for co-located multimodal edge federated learning that jointly optimizes client selection and accuracy allocation per round to maximize learning utility per unit of active time. Its key innovation lies in a Value-per-Hour Pricing mechanism, which dualizes resource constraints into closed-form client pricing, integrated with modality coverage constraints and a reverse water-filling strategy to enable efficient and fair multimodal collaboration. Evaluated on the nuScenes benchmark, FedSceneX reduces active time per round to 3.31 hours—significantly lower than baselines (4.85–9.78 hours)—achieves peak accuracy within 20 hours while preserving all four modalities, and sustains this advantage up to 45 hours.

Abstract

Federated learning at the sensing edge is typically evaluated by communication rounds, yet a round does not represent a fixed amount of work. Even on identical hardware, the methods we compare require 3.3 to 9.8 hours per round, which makes round-based comparisons misleading. The problem is more obvious for same-scene multimodal clients, since camera, video, LiDAR, and radar workloads differ substantially in training and communication cost, while existing methods treat the modality composition of each round as fixed. To address it, we introduce FedSceneX, an orchestrator that jointly determines round composition to maximize learning value per active hour. The optimization method, Value-per-Hour Pricing (VHP), converts the fractional objective through a parametric transformation and dualizes the uplink constraint, yielding a closed-form client price whose weights capture resource shadow costs. Based on these prices, FedSceneX selects clients subject to a modality coverage constraint, allocates precision through reverse water filling, and assigns updates to edge servers. On the full nuScenes benchmark with fifteen clients and twelve baselines, FedSceneX reduces the active time per round to 3.31 hours, compared with 4.85 to 9.78 hours for the baselines. Across all random seeds, it achieves the highest accuracy within a twenty-hour budget while preserving all four modalities. Its advantage persists from ten to forty-five hours, after which conventional methods overtake it.

Problem

• federated learning
• multimodal
• edge computing
• time-to-target
• resource heterogeneity

Innovation

• Federated Learning
• Multimodal Edge Learning
• Time-to-Target Optimization
• Value-per-Hour Pricing
• Resource Orchestration