In recent years, the modern battlefield is characterized by guerilla warfare based on enemy with low signature, hidden within civilian population. At the same time, the modern combat weapons (air and land systems) are equipped with a variety of sensors such as radars, optic sensors, sigint sensors, thermal sensors etc.
The sensors are installed on search agents such as mobile autonomous robots, vehicles or drones. These sensors obtain information about unusual events (firing, movements etc.). However, the noisy environment creates a large number of false alarm events.
The information which is collected over the time by the variety of sensors is significant. It allows filtering the false alarm events and detects the real targets, automatically in a short time.
These capabilities are achieved by advanced models, using various `Bayesian` approaches. These models include real-time updating of the location probabilities of the targets within the search area; integrating the various sensors’ outputs to generate a central ‘location-probability map’, and integrating various agents’ information, under different information-sharing policies.
This research deals with the construction of autonomous targets’ detection system within large search areas. The system is based on search algorithms and on decision-making methods using artificial intelligence and machine learning tools.
Several algorithms are developed and analyzed, including the following control policies:
Currently, a software simulation is being developed to analyze the different policies and initial results, which compare the different algorithms, are studied.