Resource-aware fusion in multimodal sensing under network constraints

Due to insufficient network infrastructure, tactical networks suffer the imbalance between network resource supply and transmission demands generated by deployed sensing devices. On top of that, those sensing devices are in multiple modalities with the need of fusing diverse inputs into unified formats to provide answers to query tasks. We addressed issues under network resource constraints on exploiting multiple distinctive sensing modalities. Then, we proposed a fusion framework to aggregate various types of sensing inputs, and an updating algorithm to adaptively schedule resource, e.g. bandwidth, to most contributing devices. With the adjustment of mapping functions which get bandwidth consumption bounded by the threshold, our algorithm keeps updating the combination of bandwidth shares to suit for streaming events as well as different query tasks. Taking multimodal sensing into account, our elementary fusion framework may shed light on deeper exploration on the correlation among variety of sensing inputs, especially under resource-constrained circumstances.