Drone Teams Now Share Heavy Loads Without Tangled Cables or Crashes

Researchers are tackling the complex challenge of coordinating multiple drones to transport heavy loads efficiently and safely. Lamberto Vazquez-Soqui, Fatima Oliva-Palomo, and Diego Mercado-Ravell, from the Center for Research in Mathematics (CIMAT AC), alongside Pedro Castillo of Université de Technologie de Compiègne, present a novel approach to distributing cable tension in multi-drone systems. Their work addresses the inherent problem of uneven energy use and potential instability when multiple drones work together to carry a single load. By integrating a Sequential Quadratic Programming (SQP) algorithm into existing control systems, the team demonstrates a significant improvement in cable management, preventing slack and collisions while maintaining accurate trajectory tracking. This optimisation layer offers a scalable solution for achieving energy-balanced and secure cooperative load transport, paving the way for practical applications of multi-agent aerial systems.

Multi-Agent Aerial Load Transport Systems (MAATS) offer significant advantages over single-drone solutions, particularly in payload capacity and fault tolerance.

However, a fundamental challenge lies in the underdetermined tension allocation problem, which can lead to uneven energy use, cable slack, and potential collisions. This penalty actively discourages small angles between cables, effectively preventing the drones from converging and maintaining a safe operating distance. Numerical simulations, conducted with four quadrotor drones, demonstrate the effectiveness of this approach under nominal conditions.
The SQP routine consistently executes in just a few milliseconds on standard hardware, confirming its feasibility for real-time implementation in practical applications. A detailed sensitivity analysis reveals that the cable-alignment penalty gain can be adjusted online. This tunability allows for a controllable trade-off between safety margins and energy consumption without compromising tracking performance in simulation.

By minimizing both tension and inter-cable angles, the research provides a scalable framework for safe and energy-balanced cooperative load transport. This innovation represents a significant step towards reliable and efficient multi-drone operations in various logistical and industrial settings. The study’s hierarchical control architecture consists of four layers: load control, tension allocation, position control, and attitude control.

The newly implemented SQP algorithm functions as the tension allocation layer, calculating individual tension magnitudes and directions for each drone based on the virtual controller’s desired total force. This feedback structure ensures that actual cable tensions remain within safe limits while accurately following the optimized references.

The system dynamics consider n UAVs transporting a point mass via rigid, massless cables of length L i *, where each UAV *i has mass *m i * and inertia matrix *J i *. The SQP formulation operates by minimising the sum of squared cable tensions, directly addressing the issue of uneven energy distribution amongst drones.

Simultaneously, it incorporates a cable-alignment penalty that actively discourages small angles between cables, preventing convergence and potential collisions without altering the intended payload trajectory. Numerical simulations utilising four quadrotor models were conducted to test the method under nominal operating conditions.

These simulations employed standard hardware to assess the computational performance of the SQP routine, demonstrating execution times consistently within a few milliseconds. This rapid processing speed confirms the feasibility of implementing the optimisation layer for real-time applications in dynamic environments.

The study further involved a sensitivity analysis focused on the cable-alignment penalty gain. This analysis revealed that the gain can be tuned online, allowing for a controllable trade-off between maintaining a safe operational margin and minimising energy consumption. Importantly, this tuning process exhibited no measurable degradation in tracking performance throughout the simulations.

The research implemented this SQP layer to augment an existing hierarchical control structure, explicitly addressing both tension balancing and cable alignment. The system, consisting of four quadrotors, maintained a root mean square (RMS) position error of 2.97cm while transporting a 0.225kg payload via 1.0m cables over a 20s mission.

Maximum positional excursion never exceeded 8cm, matching the accuracy of previous multi-agent aerial transport studies that relied on simpler force allocators. The SQP allocator achieved a significantly more balanced force distribution compared to a baseline geometric pattern approach, as demonstrated by cable tension profiles.

The baseline controller exhibited uneven tension, while the SQP-based allocation restricted tension fluctuations to a small range, ensuring more uniform battery discharge. The alignment penalty maintained all pairwise cable angles comfortably above a 30◦ safety threshold throughout operation, with the worst-case minimum angle reaching 44.4◦.

Total simultaneous tension peaked at 3.33 N, 1.24times its long-term average, indicating stable force management. Integrated tension cost, a measure of cumulative cable force, varied between 26.8 and 33.7 Nůs depending on the alignment weight, with minimum inter-cable angles ranging from 44.4◦ to 47.9◦. A complete SQP cycle averaged 1.13ms on a Windows 11 laptop with an Intel Core Ultra 7-155U processor, with 99th percentile and worst-case times of 3.06ms and 12.29ms respectively, confirming real-time feasibility.

Real-time cable tension minimisation for cooperative multi-rotor load transport

Researchers have developed a real-time optimization layer for multi-agent aerial load transport systems, addressing challenges related to energy distribution and cable management. The system achieves balanced tension distribution amongst the drones while adhering to specified safety parameters, such as maintaining appropriate angles between cables.

Computational experiments reveal that the optimization routine operates within a few milliseconds on conventional computing hardware, confirming its suitability for real-time applications. A sensitivity analysis further indicates that the system’s performance can be tuned to balance safety margins and energy consumption without compromising trajectory tracking accuracy in simulated environments.

This framework therefore offers a scalable solution for safe and energy-efficient cooperative load transport in practical settings. The authors acknowledge that the current work relies on simulations and assumes ideal conditions, such as rigid cables and point-mass payloads. Future research will likely focus on validating the approach with physical prototypes and addressing the complexities introduced by real-world factors like cable elasticity, aerodynamic disturbances, and sensor noise. Further investigation into adaptive tuning of the cable-alignment penalty based on environmental conditions and payload dynamics could also enhance the robustness and efficiency of the system.