In terrestrial wireless networks, base stations are anchored to the ground. Network optimization is purely a resource allocation problem: how to distribute power, bandwidth, and time slots across a static spatial landscape. However, when base stations take flight—mounted on Unmanned Aerial Vehicles (UAVs) or robotic platforms—the physical topology of the network becomes a highly controllable degree of freedom. This paradigm shift introduces a deep coupling between the physics of flight and the mathematics of information theory.
As highlighted by Yong Zeng 0001 et al. in their Guest Editorial and further emphasized by Jianchao Zheng et al. in Guest Editorial: Aerial Computing: Drones for Multi-Access Edge Computing, treating an aerial node's trajectory and its communication parameters as independent problems leads to severe operational inefficiencies. Joint trajectory and resource optimization treats physical mobility and radio resource allocation as a single, unified optimization problem to maximize network performance and energy efficiency.
The Core Tension: Propulsion vs. Transmission Energy
To understand why joint optimization is necessary, consider the fundamental energy trade-off of an aerial node. Drones are strictly battery-constrained, and the vast majority of their energy is consumed by mechanical propulsion to stay airborne and move. Wireless transmission also consumes energy, which scales non-linearly with distance due to path loss.
If a drone plans its flight path purely to minimize travel distance, it may remain far from its users, forcing it to expend massive transmission power or suffer low data rates. Conversely, if it flies directly over every user to maximize signal strength, it wastes precious propulsion energy on extended flight paths.
The solution is to synchronize the spatial path with the temporal demands of the network. For instance, Meng Hua et al. demonstrate in Energy Optimization for Cellular-Connected Multi-UAV Mobile Edge Computing Systems with Multi-Access Schemes that when a trajectory is optimized jointly with task offloading schedules, the aerial node dynamically dips closer to ground stations during peak data transmission periods. This spatial adjustment minimizes transmission power exactly when data requirements are highest, balancing the energy equations of both flight and communication.
In extreme cases, such as the laser-powered UAVs studied by Jie Ouyang et al. in Throughput Maximization for Laser-Powered UAV Wireless Communication Systems, this joint optimization allows the vehicle to navigate through energy-harvesting beams to fully offset both propulsion and communication costs, achieving sustainable flight.
The Fallacy of Decoupled Optimization
Why can't we just optimize the path first and the communication resources second? Because the vehicle's position dictates the channel state, which in turn dictates resource efficiency.
Hongliang Zhang et al. point out in Cooperation Techniques for A Cellular Internet of Unmanned Aerial Vehicles that independent optimization forces mobile agents to make sequential, sub-optimal choices—such as pausing data collection entirely to relocate for better signal reception. By contrast, integrating flight path planning with signal strength management allows the system to balance the proximity required for sensing against the connectivity needed for data transmission simultaneously.
Similarly, Amine Bejaoui et al. show in A QoS-Oriented Trajectory Optimization in Swarming Unmanned-Aerial-Vehicles Communications that jointly optimizing aerial trajectories, power allocation, and user scheduling provides massive throughput gains compared to static or simple circular paths, preventing the fairness imbalances that occur when path planning ignores real-time interference.
Kinematic Patterns: The Hover-Fly-Hover Strategy
When algorithms optimize these joint problems, they rarely produce smooth, continuous curves. Instead, they often converge on a distinct kinematic pattern: the hover-fly-hover strategy.
Kaitao Meng analyzes this behavior in UAV Trajectory and Beamforming Optimization for Integrated Periodic Sensing and Communication, demonstrating that the optimal trajectory alternates between stationary periods at mathematically ideal locations and maximum-speed transitions between them. By hovering at the sweet spot where the Signal-to-Noise Ratio (SNR) is maximized for a specific cluster of ground users, the drone maximizes its data transfer rate. Once the transmission or sensing task is complete, it flies as fast as possible to the next optimal location, minimizing the time spent in low-utility transition zones.
As Junbin Cui et al. note in Joint trajectory design and resource allocation for UAV-assisted mobile edge computing in power convergence network, this strategy naturally prioritizes flying closer to terminals with larger data payloads, directly translating spatial proximity into reduced latency.
How Advanced Wireless Technologies Reshape Flight
The coupling between trajectory and resources becomes even more pronounced when advanced physical-layer technologies are introduced, as they fundamentally alter the geometric constraints of the flight path.
Non-Orthogonal Multiple Access (NOMA)
In traditional orthogonal access, users get dedicated time or frequency slots. NOMA allows a base station to serve multiple users simultaneously on the same frequency by using superposition coding and successive interference cancellation. Yuanwei Liu et al. in UAV Communications Based on Non-Orthogonal Multiple Access and Xidong Mu et al. in Non-Orthogonal Multiple Access for Air-to-Ground Communication show that optimizing trajectories alongside NOMA power allocation ensures that signal decoding priorities remain valid even as the physical distance between the transmitter and receivers changes dynamically.
Remarkably, Mushu Li et al. discovered in Energy-Efficient UAV-Assisted Mobile Edge Computing: Resource Allocation and Trajectory Optimization that NOMA causes optimal flight trajectories to contract. Because NOMA increases the overall channel capacity, the aerial server can maintain required data rates from a greater distance, allowing it to shorten its flight path and conserve substantial propulsion energy.
Reconfigurable Intelligent Surfaces (RIS)
RIS technology uses smart, programmable surfaces to reflect wireless signals and bypass environmental blockages. When combined with mobile robots or drones, the trajectory design expands to include the environment itself. Wanli Ni et al. in Reconfigurable Intelligent Surface for Internet of Robotic Things and Xinyu Gao et al. in Trajectory and Passive Beamforming Design for IRS-aided Multi-Robot NOMA Indoor Networks demonstrate that coordinating physical movement with RIS reflection coefficients allows mobile agents to navigate along paths that maximize signal quality and energy harvesting, proving that path planning based on environmental signal availability is far more efficient than simply boosting raw transmit power.
Information-Centric and Multi-Agent Coordination
Modern formulations of this problem extend beyond raw throughput to incorporate information-centric metrics like data freshness and semantic priority.
For instance, Mengjie Yi et al. focus on minimizing the Age of Information (AoI) in Deep Reinforcement Learning for Fresh Data Collection in UAV-assisted IoT Networks. By allowing a reinforcement learning agent to select both a directional flight action and a specific data source in each time step, the vehicle's path is dynamically driven by the immediate data freshness needs of the network rather than geometric distance. When dealing with complex data like images, Jiawen Kang 0001 et al. show in Personalized Saliency in Task-Oriented Semantic Communications: Image Transmission and Performance Analysis that integrating semantic priority with trajectory planning is essential to mitigate large-scale fading over dynamic distances.
Finally, when multiple aerial nodes operate simultaneously, the optimization must account for network topology and mutual interference. As Irshad A. Meer argues in Reinforcement Learning Based Dynamic Power Control for UAV Mobility Management, joint management of flight paths and transmission power is required to maximize energy efficiency in complex, spatially-varying environments.
To handle this decentralized complexity, Shimin Gong et al. demonstrate in Bayesian Optimization Enhanced Deep Reinforcement Learning for Trajectory Planning and Network Formation in Multi-UAV Networks that physical trajectories and network topology are inextricably linked; changes in data relaying strategies must immediately trigger flight path adjustments. Multi-agent reinforcement learning frameworks, such as those proposed by Ursula Challita et al. in Interference Management for Cellular-Connected UAVs: A Deep Reinforcement Learning Approach, enable autonomous agents to simultaneously learn movement paths, transmit power, and cell associations, balancing localized latency requirements against global network interference.
Key Foundational Work
Communications and Control for Wireless Drone-Based Antenna Array (2017) — contributed 2 concept elements to this direction.
Multi-Node Wireless Energy Charging in Sensor Networks (2015) — contributed 1 concept elements to this direction.