The Intelligent Prospector v1.0: geoscientific model development and prediction by sequential data acquisition planning with application to mineral exploration

[摘要] Geoscientific models are based on geoscientific data; hence,building better models, in the sense of attaining better predictions, oftenmeans acquiring additional data. In decision theory, questions of whatadditional data are expected to best improve predictions and decisions is withinthe realm of value of information and Bayesian optimal survey design.However, these approaches often evaluate the optimality of one additionaldata acquisition campaign at a time. In many real settings, certainly inthose related to the exploration of Earth resources, a largesequence of data acquisition campaigns possibly needs to be planned. Geoscientificdata acquisition can be expensive and time-consuming, requiring effectivemeasurement campaign planning to optimally allocate resources. Eachmeasurement in a data acquisition sequence has the potential to inform wherebest to take the following measurements; however, directly optimizing aclosed-loop measurement sequence requires solving an intractablecombinatoric search problem. In this work, we formulate the sequentialgeoscientific data acquisition problem as a partially observable Markovdecision process (POMDP). We then present methodologies to solve thesequential problem using Monte Carlo planning methods. We demonstrate theeffectiveness of the proposed approach on a simple 2D synthetic explorationproblem. Tests show that the proposed sequential approach is significantlymore effective at reducing uncertainty than conventional methods. Althoughour approach is discussed in the context of mineral resource exploration, itlikely has bearing on other types of geoscientific model questions.

[学科分类] 土木及结构工程学