Routing in Space-Air-Ground Integrated Networks (SAGIN) demands multi-commodity optimization over fluid 3D topologies. Shortest-path heuristics treat links independently, incurring overlapping capacity penalties. Conversely, dense combinatorial formulations waste variables on edges outside communication range. We introduce Sparse-QUBO-Route, a QuantumInspired routing framework. Per epoch, it prunes distanceinfeasible edges using a Conformal Geometric Algebra (CGA) predicate. It encodes the remaining multi-commodity problem as a temporally regularized Quadratic Unconstrained Binary Optimization (QUBO) instance, which is solved by a warmstarted Quantum-Inspired Evolutionary Algorithm (QiEA). The framework is evaluated on a synthetic SAGIN testbed against OSPF, capacity-aware CSPF, and an unpruned Vanilla QUBO under a common True Cost metric, on grids up to N=50 nodes at K=2 and K=5 concurrent flows at N=26, with an exact bruteforce/MILP reference at N=5. Geometric pruning removes 25– 53% of QUBO variables, yielding 1.57×–2.72× runtime speedups (up to ≈ 3.3× at RUAV=40 km). Under bottleneck congestion (Ce=2), Sparse-QUBO-Route attains markedly lower True Cost than capacity-unaware OSPF (+12.15±2.32 vs. +40.51; 10 seeds, Wilcoxon p=0.002). However, capacity-aware CSPF achieves the lowest cost (−10.85 ± 2.63). Three findings are negative: warmstarting fails to improve route stability, the CGA filter is 7.9× slower per call than a Euclidean equivalent, and the QUBO advantage over OSPF vanishes on a less-congested multi-LEO topology. Pruning primarily buys runtime rather than route quality. The QUBO benefit is congestion-dependent, scoping Sparse-QUBO-Route to single-bottleneck backhauls where a capacity-aware sequential baseline is unavailable.

Space-Air-Ground Integrated Networks (SAGIN),Quadratic Unconstrained Binary Optimization (QUBO),Quantum-Inspired Evolutionary Algorithm (QiEA),conformal geometric algebra (CGA),multi-commodity routing