Organizations: Department of Mechanical and Aerospace Engineering, The Hong Kong University of Science and Technology, Hong Kong, China · Institute of Intelligent Transportation Systems, College of Civil Engineering and Architecture, Zhejiang University, Hangzhou 310058, China
Unmanned aerial vehicles (UAVs) are increasingly deployed in low-altitude urban airspace for delivery, emergency response, and other applications. However, UAV mobility across spatially heterogeneous urban communication environments can lead to time-varying communication outages, disrupting mission execution and compromising flight safety. To address this challenge, this paper proposes a Spatio-Temporal Relay Cooperation and Planning (ST-RCP) framework that leverages delivery UAVs as opportunistic communication relays for disconnected UAVs without requiring additional communication infrastructure. The framework predicts communication blind-spot events over a finite planning horizon, constructs a helper-event compatibility graph, and jointly optimizes relay selection, obstacle-aware trajectory replanning, and communication-aware scheduling. Multiple planners are evaluated to investigate the trade-offs among communication coverage, computational cost, and energy consumption. Experiments based on real-world traffic scenarios compare the performance of ST-RCP against baseline methods. In the peak scenario, ST-RCP improves descent-phase communication coverage by 29.93% compared with the base-station baseline and by 27.38%, 22.02%, and 15.20% compared with configurations using one, two, and three dedicated relay UAVs, respectively. Process flexibility analysis demonstrates that predictive task-aware relaying expands the pool of feasible relay resources and improves communication continuity while maintaining the feasibility of delivery missions. These findings demonstrate the potential of trajectory-enabled UAV coordination to enhance the safety and resilience of urban aerial operations.
Figures & tables
Fig. 1 : Urban canyons disrupt communication links during urban delivery task.
Fig. 2 : Overall closed-loop framework of ST-RCP. Delivery demand, communication and obstacle maps, and UAV/mission states provide the system inputs. Future connectivity degradation is converted into time-bounded blind-zone events, which are matched with eligible helper UAVs through a helper–event bipartite graph and relay cooperation. Selected helper–group assignments are then processed by constraint-aware Bézier trajectory replanning to generate feasible relay trajectories. The updated UAV and mission states are fed back to the next planning epoch, enabling continuous spatio-temporal relay coordination.
Fig. 3 : Urban simulation environment and communication maps: (a) reconstructed three-dimensional environment of Kwun Tong, Hong Kong; (b) Stepped cut of the 3-base-station SNR field; and (c) Preplanned UAV delivery trajectories over four altitude layers.
Fig. 4 : Delivery scenarios used in the main evaluation.
Parameter
Setting
Horizontal map size
1600×1400m
Voxel/radio resolution
1m
Radio altitude range
0 – 150m
Main base-station count
5
Sensitivity base-station count
3
Base-station altitude
10m
TABLE I : Experimental parameter settings.
Method
Total Coverage
Descent Coverage
Total Energy (J)
Additional Energy (J)
Base Station
82.50%
24.62%
55187.44
0.00
FANET
89.15%
53.64%
55187.44
0.00
Dense Base Station (8)
91.63%
78.01%
55187.44
0.00
Dedicated Relay UAV (3)
95.63%
81.86%
62135.56
6948.12
ST-RCP
92.27%
66.86%
56304.26
1116.82
TABLE II : Performance comparison in the 20-order off-peak scenario.
Method
Total Coverage
Descent Coverage
Total Energy (J)
Additional Energy (J)
Base Station
76.62%
51.73%
346691.87
0.00
FANET
93.72%
63.16%
346691.87
0.00
Dense Base Station (8)
90.08%
68.71%
346691.87
0.00
ST-RCP
93.63%
81.66%
359134.82
12442.95
Dedicated Relay UAV (1)
90.57%
54.28%
359407.28
12715.41
Dedicated Relay UAV (2)
91.81%
59.64%
363727.20
17035.33
TABLE III : Performance comparison in the 90-order peak scenario.
Optimizer
Objective
Coverage (%)
Descent Coverage (%)
Disconnect Count
Avg. Outage (s)
Additional Energy (J)
Greedy
Target Coverage
90.77
79.16
152
38.26
16253.95
Network Coverage
91.30
77.30
161
36.24
18843.32
Coverage per Energy
90.83
75.80
204
38.54
19576.21
Coverage per Distance
91.48
77.07
197
35.59
19713.34
GA
Target Coverage
91.56 ± 0.36
79.24 ± 1.38
138.0 ± 23.8
35.45 ± 1.42
16367.47 ± 1394
Network Coverage
93.36 ± 1.07
76.90 ± 1.26
125.3 ± 7.1
27.15 ± 4.63
14501.68 ± 888
TABLE IV : Communication-oriented objective sensitivity analysis.
Fig. 5 : Sensitivity and computational analysis of ST-RCP: (a) impact of base-station density; (b) impact of relay-following distance; and (c) optimization time of ST-RCP and dedicated-relay strategies.
Fig. 7 : Spatiotemporal evolution of a proactive relay episode under the 90-order peak scenario at t=503 , 563 , 623 , and 741.5 s. The helper UAV (blue) proactively establishes and maintains the relay link (green) for the blind-zone UAV (orange) within the coverage hole, and subsequently resumes its nominal delivery route upon task completion.
Department of Mechanical and Industrial Engineering, University of Illinois Chicago, IL, 60607 USA · DEVCOM Army Research Laboratory, Minneapolis, Minnesota, USA · DEVCOM Army Research Laboratory, Aberdeen Proving Grounds, Aberdeen, MD 21005 USA