LAUSR

The centralized computing model in industrial Internet of Things (IIoT) leads to large delay and unbalanced traffic, which strictly restricts the adoption of IIoT in industrial applications demanding high network performance. To cope with the problem, researchers resort to the in-network computation model in which the computation capability is distributed among nodes in IIoT. Existing works on the in-network computation assume that the network connectivity built in advance meets the performance requirement of the in-network computation model. Nevertheless, no node placement methods have been proposed to build network connectivity supporting in-network computation. For this reason, we propose an in-network-computation-oriented node placement (INP) algorithm. The INP algorithm first decomposes the whole problem into several delay constrained relay node placement problems, and then, solves them sequentially. Moreover, we prove that the INP algorithm ensures explicit time complexity and approximation ratio. Finally, we verify the efficiency of this work through extensive simulations and preliminary experiments.