<p>This paper investigates efficient path planning for aerial robots engaged in searching for and tracking multiple moving targets using superposition measurements. To achieve this, the multi-target states are modeled as a labeled multi-Bernoulli random finite set, enabling the estimation of both the number of targets and their states while tracking their trajectories. The path-planning challenge is formulated within the framework of a partially observed Markov decision process. The primary objective is to enhance multi-target tracking performance by guiding the aerial robots along optimal paths that maximize an information-driven reward function, represented by the Rényi divergence. This metric quantifies the information gain between labeled multi-Bernoulli prior and posterior densities, effectively increasing the observability and information quality of received signals. Beyond improving tracking efficacy, the proposed path-planning approach also addresses critical operational constraints, including minimizing fuel consumption and ensuring collision-free navigation. This not only includes avoiding collisions between aerial robots, but also avoiding obstacles or potential threats in their environment. While the multi-Bernoulli filter is known for its computational efficiency and accuracy in multi-target tracking, it lacks a direct analytical solution. To overcome this, a sequential Monte Carlo method is introduced. Numerous simulations were conducted using aerial robots equipped with low-cost received signal strength indicator sensors to test various multi-target track-before-detect scenarios. The results show a very effective synergy between the tracking filter and the path-planning module, especially in conditions of low signal-to-noise ratio. This interaction emphasizes the robustness and practical value of the proposed approach in challenging environments.</p>