Contents I. Introduction
Database assisted dynamic resource allocation is generally considered as a technique to enable network level deployment of cognitive radios [1]–[3]. Such a database ideally should include all the required information of the incumbent network (e.g., power, location, radiation pattern, bandwidth, direction of transmission, etc.) for the cognitive system intending to share the same spectrum as incumbent users, to be able to adapt its transmission parameters to the environment, without hindering operation of incumbent users. Most of the databases are obtained by collecting information from the regulatory bodies. However, such information are either not complete, or becomes outdated after a short time. This calls for a dynamic technique in order to complete the information of databases, update the existing information, or even produce a database where such information can not be obtained from regulatory bodies. Spectrum cartography or radio environment mapping is proposed as an efficient technique to produce the dynamic database of the incumbent or primary users, [4]–[8]. However, spectrum cartography can have plethora of other applications, e.g., network monitoring, malicious user detection, interference monitoring, and etc. The cornerstone of any spectrum cartography technique is a collaboration of sensors to estimate source parameters, e.g., location and power [9]–[15]. Bazerque and Giannakis [9] employ sparse signal processing techniques to localize and estimate the power of multiple incumbent transmitters. In [10], quantized measurements are used to reduce the communications overhead and overcome the hardware complexities. And, location of incumbent users are determined in [11] assuming a fading channel model. Most of these works provide efficient tools for spectrum cartography of omni-directional sources which can be a valid assumption for lower parts of the frequency spectrum. However, considering the highly directive nature of wireless communications in higher parts of spectrum (e.g., Ka band, mmWave, etc. [16]), estimation of direction of transmission (DoT) becomes an essential component of spectrum cartography in order to obtain accurate results. For example, terrestrial microwave links in Ka band often used for mobile backhauling are highly directive, and thus for the cognitive systems such as fixed satellite services to coexist with the terrestrial links, it is important to know in which directions, the terrestrial links are operating [3], [16]. The same holds when a new terrestrial system intends to reuse the frequency of currently in use microwave links, e.g., for smart backhauling [17]. In such cases, the cognitive system needs to have a good estimate of the amount of power in a specific place in order to operate properly, and determine its transmission parameters such as carrier, power, etc., [3]. Even if the cognitive system is aware of all the underlying parameters, e.g., source power, location, etc., but still the knowledge of DoT is essential. Otherwise, the cognitive system is not able to obtain an accurate estimate of the power distribution in the environment, and either may hinder the operation of incumbent users or adapt transmission parameters which are not efficient.
