*Result*: Geographic-based failure-tolerant routing for LEO mega-constellations.

*The rapid development of low Earth orbit (LEO) mega-constellations has provided significant opportunities for achieving global network services with wide coverage, low latency, and high bandwidth. However, the orbital dynamics of satellites lead to frequent changes in network topology, and the complex operating environment poses significant challenges for achieving reliable routing transmission. Geographic routing algorithms, as low-complexity distributed algorithms, are effective in addressing the dynamic nature of network environments. However, when applied to LEO satellite networks, these algorithms face several limitations, with the routing void problem being the most critical. To address this, this paper proposes a Locally Aware Geographic Routing (LAGR) algorithm, which ensures normal greedy routing operation by perceiving an expanded network scope. Furthermore, to enhance LAGR's robustness, we propose FT-LAGR, which strengthens routing resilience by sorting nodes within the local region in ascending order of proximity to the destination and iteratively checking path connectivity to these nodes. Simulation results demonstrate that FT-LAGR significantly outperforms baseline algorithms across latency, packet loss rate, and throughput metrics. Notably, under a 10% link failure probability, FT-LAGR achieves a packet loss rate reduction exceeding 40%, throughput more than twice, and latency reduction to less than one-third compared to algorithms equipped with packet retransmission mechanisms. [ABSTRACT FROM AUTHOR]*