A Human-Automation Interaction Approach to the Evaluation of Resource Allocation Strategies in Adaptive Distributed Sensor Networks

How to design and to evaluate adaptive distributed sensor networks with respect to end user decision making is an important topic in human-automation interaction. This study evaluates four resource allocation strategies (static automation, adaptive sensing, mixed-initiative control, and control incorporating end user task analytic information) in a simulated prototype weather sensor network. Performance of each sensor network resource allocation strategy was assessed with measures of sensor coverage quality over a target geographic region of interest. Coverage measures calculated from the number of radar nodes, tilts and cycles were derived for partial (radar scan touches at least part of the analysis region) and complete coverage of the region by single and multiple (two or more) radars. Single and multi-radar coverage area scores were also derived from the cumulative fraction of the analysis regions area that was scanned. Using three weather scenarios, results show that static automation significantly outperforms the other strategies in all cases using measures of single and multi-radar complete coverage and for a uniformly distributed weather event using single and multi-radar coverage scores. Otherwise the three strategies outperform static automation. As the performance of resource allocation strategies are sensitive to the weather events themselves as well as the measures used to evaluate them, the design of next generation of algorithms should consider these performance distinctions.