Abstract: With the full deployment of the BeiDou-3 Navigation Satellite System (BDS-3), the number of visible satellites per epoch routinely exceeds 30. The substantial volume of GNSS observational data presents a severe challenge to the on-chip resources and computing power of consumer-grade terminals, such as low-altitude unmanned aerial vehicles. To address this issue, a fast satellite selection algorithm based on the improved beetle antennae search (IBAS) is proposed. Building upon the standard modified beetle antennae search (MBAS), this method incorporates a weighted directional search strategy and a hybrid stopping mechanism to enhance global search capability and convergence efficiency. Semi-physical simulations utilizing domestic low-power chips demonstrate that, compared to the MBAS algorithm, IBAS reduces the average execution time for satellite selection by 60%-65%. When selecting 12 satellites, the algorithm achieves a 100% capture rate for optimal geometric configurations. Furthermore, it maintains sub-meter positioning accuracy even when only 6 satellites are retained for calculation. Consequently, the proposed method significantly reduces computational overhead while ensuring positioning reliability, rendering it highly suitable for real-time deployment on resource-constrained terminals.