With the unique advantages that very low Earth orbit (VLEO) satellites have demonstrated in fields such as high-resolution remote sensing, the contradiction between the demand for large-volume data downlink and the extremely short visibility windows of ground stations has become increasingly prominent. Sun-synchronous orbit VLEO satellites typically have only one to two ground station passes per day, each lasting only a few minutes, far from sufficient to meet the high-throughput data downlink requirements generated by missions such as high-resolution imaging and video surveillance. However, existing relay architectures are unsuitable for small satellites due to constraints in mass, power consumption, and latency; low-rate terminals can only provide basic telemetry; conventional proportional–integral–derivative controllers struggle to maintain accuracy under nonlinear and strong disturbance environments; and nonlinear model predictive control imposes excessive computational burdens. Therefore, how to coordinately optimize the resource allocation and control strategies of VLEO satellites and relay constellations to achieve an effective balance between data transmission efficiency and onboard computational capability has become a key technical challenge in supporting high-capacity VLEO missions.
In a recent study published in Space: Science & Technology , researchers from Nanyang Technological University, Singapore, proposed a bi-objective optimization and control framework for VLEO satellites based on a data-relay multi-satellite system. The study constructs a cooperative network consisting of one VLEO transmitting satellite and multiple relay satellites. The transmitting satellite offloads data to the relay satellites during mutually visible windows, and the relay satellites cache the data and forward it during ground station overpasses, thereby extending the downlink window. The transmitting satellite adjusts its attitude via particle swarm optimization to maximize solar energy intake, while the relay satellites maximize the communication gain toward the ground station while maintaining alignment with the transmitting satellite. On this basis, a hybrid control framework based on asynchronous multi-model coordination is proposed: the transmitting satellite employs conventional model predictive control to achieve high-precision power optimization and disturbance rejection, while the relay satellites employ multiplexed model predictive control to sequentially update actuator commands in an interleaved manner, significantly reducing computational load. Hardware-in-the-loop simulation results demonstrate that the proposed method outperforms conventional model predictive control and differential evolution-based comparison schemes in terms of attitude control accuracy for both relay satellites and the transmitting satellite, with the pointing errors of the relay satellites toward the ground station and the transmitting satellite both constrained within allowable limits. This study provides a collaborative scheme that balances computational feasibility and control precision for high-capacity data relay of VLEO satellites, offering significant engineering application value for the construction of next-generation high-throughput, low-power VLEO satellite relay communication systems.
First, this paper focuses on the communication bottleneck problem faced by high-capacity data downlink of very low Earth orbit (VLEO) satellites and proposes a data-relay multi-satellite system architecture. When operating at an altitude of 300 km, VLEO satellites generate massive data from their high-resolution imaging payloads; however, constrained by the extremely short ground station visibility windows, over a simulation period of 10 days, the total ground contact time for a satellite at 300 km altitude at 0° elevation angle is only approximately 208 minutes, representing a reduction of 43% compared with that at 550 km altitude, with each pass averaging less than 7 minutes—far from sufficient to meet the high-throughput data downlink requirements. If a wind-facing attitude is adopted to reduce the atmospheric drag area, the contact time further drops to less than 1 minute, rendering direct-to-ground communication nearly infeasible. To address this, the study proposes a data-relay multi-satellite system architecture as illustrated in Fig. 1, consisting of one VLEO transmitting satellite and multiple relay constellation satellites forming a collaborative network. During mutually visible windows, the transmitting satellite offloads data to the relay satellites, which then forward the data during their own ground station overpasses, thereby effectively extending the downlink window. Each satellite runs a local control module to autonomously manage its own state, exchanging only critical state variables—such as buffer load, line-of-sight alignment, and energy levels—through minimal inter-satellite links, with control loops executed asynchronously, enabling cooperative operation without the need for centralized scheduling.
Second, this paper establishes the dynamic model of the VLEO satellite and proposes a bi-objective optimization strategy based on particle swarm optimization (PSO) and a hybrid model predictive control framework based on asynchronous multi-model coordination. Owing to the presence of large solar panels, VLEO satellites are subjected to significant disturbance torques—including atmospheric drag, geomagnetic disturbances, and solar panel-induced perturbations—resulting in highly nonlinear attitude dynamics. As illustrated in Fig. 2, the control architecture assigns different tasks according to satellite roles: the transmitting satellite employs conventional model predictive control (MPC) to achieve power optimization over a short control horizon, adjusting its attitude to maximize the solar panel illumination area while maintaining antenna alignment with the relay satellites. Its cost function simultaneously incorporates pointing error and solar incidence angle, with the latter assigned a substantially higher weight to prioritize electrical power acquisition. The relay satellites, in contrast, employ multiplexed model predictive control, with a cost function that prioritizes precise pointing toward the ground station to ensure downlink availability, while constraining the pointing angle toward the transmitting satellite within allowable limits. Both types of satellites adopt the particle swarm optimization algorithm to solve for the optimal attitude configuration at each control cycle. Compared with gradient descent methods and differential evolution algorithms, PSO exhibits faster convergence and is less prone to local optima when handling such nonlinear multi-objective optimization problems. The transmitting satellite employs conventional MPC that synchronously updates control inputs across all three channels to ensure accuracy, whereas the relay satellites adopt multiplexed MPC, which sequentially updates individual channels in an interleaved manner, substantially reducing computational load.
Finally, this paper validates the effectiveness of the proposed control framework through hardware-in-the-loop (HIL) simulations. Fig. 3 presents the HIL experimental platform, which comprises physical components including reaction wheels, magnetorquers, and an attitude determination and control computer, with a sensor simulator providing realistic input data and an independent PC simulating satellite attitude dynamics, disturbance torques, and orbital motion. Fig. 4 presents the attitude control results of the relay satellite under the multiplexed/conventional multi-model predictive control combined with particle swarm optimization (MPC+PSO) method, showing that the Euler angle errors are significantly smaller than those of the two comparison schemes, namely conventional MPC+PSO and conventional MPC with differential evolution (MPC+DE). Fig. 4C further demonstrates that the convergence speed of PSO is markedly faster than that of DE, which exhibits significant oscillations during the optimization process. Figs. 5A and 5B present the pointing errors of the relay satellite toward the ground station and toward the VLEO transmitting satellite, respectively, both of which are strictly constrained within allowable limits, verifying the effectiveness of the bi-objective optimization framework in simultaneously maintaining both communication links. Fig. 6 presents the attitude control results of the transmitting satellite under the same control framework, which likewise outperforms the conventional MPC+DE scheme, with PSO exhibiting faster convergence. The HIL simulation results demonstrate the comprehensive advantages of the proposed method in terms of improved attitude control accuracy, optimized convergence speed, and relay reliability, providing a collaborative solution that balances computational feasibility and control precision for high-throughput VLEO missions.
Space Science & Technology
Dual Optimization and Control of VLEO Relay Systems via Multiplexed/Classical Multimodel MPC
24-Jul-2026