People frequently face preference-based planning tasks requiring balancing goals with nuanced constraints, yet even advanced LLMs demand considerable effort to produce and adjust plans reflecting complex user preferences. We present MAVIS (Multi-Agent Virtual Interactive Synergy), a multi-agent system within a virtual workspace that introduces an incremental collaboration mechanism. This mechanism automatically decomposes tasks into guidelines and sequentially introduces expert agents. Each agent proactively engages users in focused dialog to uncover implicit preferences, while successive agents add perspectives and transparently negotiate trade-offs. To mitigate textual overload, MAVIS employs spatial visualizations that externalize agents’ reasoning through step-linked summaries and context-aware boards, with embodied avatars supporting natural interaction. Across studies, Study 1 showed our collaboration mechanism doubled expressed preferences and improved planning quality by 60.3% over a conventional LLM baseline. Study 2 affirmed visualization’s benefits over a non-spatial baseline, while Study 3 confirmed its versatility across VR and desktop modalities and diverse tasks.

Accurate synchronization of distributed sensor measurements is essential for reliable vibration-based structural health monitoring (SHM). Wireless Internet of Things (IoT) systems, while attractive for large-scale SHM deployments, inevitably introduce unknown time lags arising from clock drift and communication delays. Conventional frequency-domain (FD) estimators and feature-based machine-learning (ML) models struggle to handle such asynchronous measurements, especially when delays are smaller than the sampling interval. To address these limitations, this paper proposed a learning-based synchronization framework that integrates a time-lag convolutional neural network with attention (TLCA) to directly estimate continuous-valued time lags from raw vibration signals, together with a resampling-based alignment strategy to correct non-integer delays. The proposed framework was validated using both simulated structural responses and field data from a wireless SHM system deployed on the Jindo Bridge. By directly learning time relationships from raw vibration signal pairs, TLCA achieves sub-millisecond accuracy and consistently outperforms conventional frequency-domain and feature-based learning methods, with about an order-of-magnitude improvement in precision. It also maintained strong robustness under increased noise and modal coupling. With efficient data-level alignment and low computational complexity of the overall framework, it can enable accurate and real-time synchronization for practical IoT-based SHM applications.