Search papers, labs, and topics across Lattice.
This paper introduces MobiWave, a novel framework that integrates a dispatch-oriented multi-scale graph wavelet module with Drift-Guided Layer-Selective Optimization (DGLS) to enhance autonomous fleet rebalancing. By effectively separating and weighting graph-frequency patterns, MobiWave improves demand prediction and rebalancing efficiency, while DGLS ensures robust adaptation to mobility drift by selectively updating model layers based on dispatch-weighted spectral drift. Experimental results on real-world datasets and simulations show that MobiWave outperforms existing state-of-the-art methods in maintaining service quality and safety while optimizing dispatch rewards.
MobiWave achieves superior fleet rebalancing by intelligently separating traffic patterns and adapting to mobility drift, outperforming traditional approaches in both efficiency and reliability.
Autonomous fleets enable mobility platforms to coordinate idle vehicles directly, making fleet-wide rebalancing possible. However, two obstacles limit reliable deployment: overlapping regional and local traffic patterns can hide roads that remain useful for dispatch, and mobility drift can make a trained policy unreliable. Existing spatial aggregation mixes these patterns, while updating all parameters from limited recent data is costly and can damage stable knowledge. We propose \name, a framework that connects a dispatch-oriented multi-scale graph wavelet module with Drift-Guided Layer-Selective Optimization (DGLS). The first module addresses the representation challenge by separating graph-frequency patterns and weighting each scale according to its value for demand prediction and feasible rebalancing. DGLS addresses the adaptation challenge by measuring Dispatch-weighted Spectral Drift, selecting affected layers within a resource budget, and separating short shocks from persistent changes through a drift-aware fast--slow update. Candidate validation further rejects updates that fail to improve held-out dispatch reward without worsening monitored service or safety constraints. Experiments on both real-world datasets and simluated environments demonstrate the effectiveness of \name\ in comparing with state-of-the-art methods. The source code and datasets are available at https://anonymous.4open.science/r/MobiWave-40F8/.
