Active logic is a time-situated reasoner that can track the history of inferences, detect contradictions, and make parallel inferences in time. In this paper, we explore the behavior of an active-logic based agent on different sets of action selection axioms for a time-constrained target search task. We compare the performance of a baseline set of axioms that does not avoid redundant actions with five other axiom sets that avoid repeated actions but vary in their knowledge content. The results of these experiments show the importance of balancing boldness and caution for target search.

As more users access the radio frequency (RF) spectrum for wireless communications, spectrum availability is becoming an increasingly scarce resource. Hence, the ability to detect or predict when a spectrum channel is available for use is of great importance. To support autonomous access to the spectrum band, we research two feature extraction techniques: (i) based on the standard energy calculation and (ii) based on cumulant calculations. We compare the performance of a baseline reactive predictor which projects the current time-step values to the next time-step, against linear support vector regression (SVR) based approaches using the aforementioned feature extraction techniques. We evaluate the occupancy state prediction in a spectrum band for different values of signal-to-noise ratio (SNR) and spectrum RF activity, using simulated RF signal data. Our experiments indicate that using first order cumulant based approach with SVR improves prediction accuracy.

This paper evaluates the ability of several machine learning algorithms to detect attacks that emulate normal nonperiodical messages in the MIL-STD-1553 communication traffic. The dataset for this research is highly imbalanced and most algorithms fail or simply produce poor results when classifying the data. We conduct several experiments with different machine learning algorithms to correctly classify MIL-STD-1553 dataset. We identify appropriate metrics to judge the performance of the models applied to this dataset. Using these metrics, we compare the performance of different machine learning models that we generated.