AbstractA mathematical model of a discrete sequential search for a target moving in discrete space is given. The model is based on a Bayesian updating algorithm giving successive probability distributions of target position at intervals throughout the search. Updating allows for target movement and for negative information gained from unsuccessful search.
The search is conducted by taking a sequence of discrete, instantaneous looks at chosen points, or nodes, of the search area. The sequence of chosen nodes is termed a strategy. The successive target position distributions allow the probability of detecting the target to be found for any strategy.
The model is an improvement over previous discrete sequential search models with respect to the following points. Target movement between nodes of the search area is formulated in terms of statistical information of target speed and direction, which are likely to be known. The time interval between looks, and target movement during this time, are related to the distance travelled by the searcher between search nodes. Also, with each look, the searcher has a view of surrounding nodes as well as the chosen search node. Implementation of these refinements is aided by considering the search area to consist of a finite, isometric pattern of nodes.
Optimisation of strategies with respect to both detection probability and detection probability per unit cost is considered, and a criterion given in each case to assist optimisation. However, in practice, these criteria are of limited use, and full optimisation can only be carried out in a limited range of cases. Restricting both the planning horizon of the optimisation process, and searcher travel distance, allows sub-optimal strategies to be found in a wider range of cases. Results suggest that the detection probability of strategies found under
these restrictions is normally close to optimal.
|Date of Award||Mar 1989|