Online Meta-Learning for Scene-Diverse Waveform-Agile Radar Target Tracking

Abstract

This paper attempts to characterize the kinds of physical scenarios in which an online learning-based cognitive radar is expected to reliably outperform a fixed rule-based waveform selection strategy, as well as the converse. We seek general insights through an examination of two decision-making scenarios, namely dynamic spectrum access and multiple-target tracking. The radar scene is characterized by inducing a statespace model and examining the structure of its underlying Markov state transition matrix, in terms of entropy rate and diagonality. It is found that entropy rate is a strong predictor of online learning-based waveform selection, while diagonality is a better predictor of fixed rule-based waveform selection. We show that these measures can be used to predict first and second-order stochastic dominance relationships, which can allow system designers to make use of simple decision rules instead of more cumbersome learning approaches under certain conditions. We validate our findings through numerical results for each application and provide guidelines for future implementations.

Linear Jamming Bandits: Sample-Efficient Learning for Non-Coherent Digital Jamming

“…(4) Update B = B + ϕ at ϕ T at , f = f + ϕ at C t , and μ = B −1 f . end that Assumption 1 is indeed pragmatic for the non-coherent jamming problem given the context features defined in (11).…”

“…Thus, more nuanced features than the expected cost of each arm are desirable for learning in the non-coherent case so that near-optimal strategies are not eliminated too quickly. Our choice of context features are seen in (11).…”

“…In (11), the latter two features capture more relevant information that aids in discerning the utility of jamming strategies, since it is assumed that very low error rates will be experienced frequently due to the unknown phase offset. This intuition is further confirmed by Figure 2, which shows the distribution of SER for the optimal jamming strategy when the victim transmits BPSK.…”

“…Remark 1: The context features listed in (11) are by no means unique, but were found to be effective for the cases tested in simulation. Additionally, these context features do not assume any access to additional information about the victim's symbol.…”

“…This has been analyzed in the context of rate-distortion theory [10], but has not been widely applied to real-world problems. Finally, since knowledge of the optimal jamming strategy for one user may provide relevant information about the optimal jamming strategy for another user, it is possible that online meta-learning and transfer learning techniques can be applied to speed up the learning process by gradually acquiring implicit information about the victims' signals [11].…”