Towards the fast and robust optimal design of wireless body area networks

Wireless body area networks are wireless sensor networks whose adoption has recently emerged and spread in important healthcare applications, such as the remote monitoring of health conditions of patients. A major issue associated with the deployment of such networks is represented by energy consumption: in general, the batteries of the sensors cannot be easily replaced and recharged, so containing the usage of energy by a rational design of the network and of the routing is crucial. Another issue is represented by traffic uncertainty: body sensors may produce data at a variable rate that is not exactly known in advance, for example because the generation of data is event-driven. Neglecting traffic uncertainty may lead to wrong design and routing decisions, which may compromise the functionality of the network and have very bad effects on the health of the patients. In order to address these issues, in this work we propose the first robust optimization model for jointly optimizing the topology and the routing in body area networks under traffic uncertainty. Since the problem may result challenging even for a state-of-the-art optimization solver, we propose an original optimization algorithm that exploits suitable linear relaxations to guide a randomized fixing of the variables, supported by an exact large variable neighborhood search. Experiments on realistic instances indicate that our algorithm performs better than a state-of-the-art solver, fast producing solutions associated with improved optimality gaps.     

Global optimization of open pit mining complexes with uncertainty

Global optimization for mining complexes aims to generate a production schedule for the various mines and processing streams that maximizes the economic value of the enterprise as a whole. Aside from the large scale of the optimization models, one of the major challenges associated with optimizing mining complexes is related to the blending and non-linear geo-metallurgical interactions in the processing streams as materials are transformed from bulk material to refined products. This work proposes a new two-stage stochastic global optimization model for the production scheduling of open pit mining complexes with uncertainty. Three combinations of metaheuristics, including simulated annealing, particle swarm optimization and differential evolution, are tested to assess the performance of the solver. Experimental results for a copper-gold mining complex demonstrate that the optimizer is capable of generating designs that reduce the risk of not meeting production targets, have 6.6% higher expected net present value than the deterministic-equivalent design and 22.6% higher net present value than an industry-standard deterministic mine planning software.