Hop-bounded informed-probabilistic roadmap for robust unmanned aerial vehicle relay placement in free-space optical networks

• Introduces a hop-bounded informed probabilistic roadmap that minimizes the number of unmanned aerial vehicle (UAV) relays while enforcing line-of-sight (LoS) and per-link free-space optical budget constraints. • Uses spatially varying atmospheric attenuation maps and a conservative operational margin to improve per-hop link reliability. • Presents a three-phase pipeline: fast single-relay feasibility check, informed probabilistic roadmap (PRM) baseline, and hop-bounded refinement on a state-expanded graph to reduce relay count. • Incorporates lightweight short-term signal prediction with exponential weighted moving average (EWMA) and Kalman filters to enable proactive replanning and shorten outage and recovery time. • Empirical evaluation on multiple urban maps and weather regimes shows about 25 to 35 percent reduction in average relay count compared to PRM and rapidly exploring random tree (RRT*) baselines while maintaining equal or lower outage probability and supporting near-real-time planning. Free space optical (FSO) links provide high capacity wireless connectivity but require strict line-of-sight (LoS) and remain vulnerable to obstacles and atmospheric attenuation. This study investigates the placement and dynamic deployment of unmanned aerial vehicle (UAV) relays for FSO networks in cluttered urban environments. The objective is to minimize the number of UAV relays while ensuring reliable LoS and link budget feasibility under spatially varying atmospheric conditions. The proposed framework applies a three phase method. The first phase performs a fast single relay feasibility test using disk intersection and visibility checks. The second phase builds a baseline solution through an informed probabilistic roadmap (PRM). The third phase introduces a modified informed PRM with hop bounding and constrained shortest path search on a state expanded graph to obtain solutions with fewer relays. The framework applies a conservative 90% operational margin and integrates spatially varying attenuation maps. It employs short term signal prediction using exponential smoothing and a Kalman filter to enable proactive re planning. Assignment and trajectory planning rely on the Hungarian algorithm, bottleneck optimization, and A* path smoothing for UAV dispatch. The evaluation on three representative urban maps and three atmospheric regimes with 100 Monte Carlo runs per case demonstrates that the proposed method reduces the average relay count by 25–35% compared with PRM and rapidly exploring random tree (RRT*) baselines while maintaining equal or lower outage probability. Predictive re planning lowers recovery time by 25–35% compared with reactive schemes. The framework improves operational cost and resilience of UAV assisted FSO networks and supports near real time deployment with moderate computation.