Communications-Aware NMPC for Multi-Rotor Aerial Relay Networks Under Jamming Interference

arXiv:2603.28467 · cs.RO · Submitted 2026-03-30 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.

Rosa: I'm Rosa, and with me are Dev and Taro, guest researcher.

Dev: Today's paper: "Communications-Aware NMPC for Multi-Rotor Aerial Relay Networks Under Jamming Interference".

Rosa: The gist:

Dev: First, who's behind it and why it matters.

Title and authors: Rosa: So we’re looking at this paper now called Communications Aware NMPC for MultiRotor Aerial Relay Networks Under Jamming Interference. It's a modular control approach that combines a high-level trajectory generator with an actuator level Nonlinear Model Predictive Controller, or NMPC.

Dev: Exactly. The authors are tackling the problem of MultiRotor Aerial Vehicles under jamming interference where losing communication stops the mission. They want to show how linking motion and communication objectives together helps improve performance significantly.

Taro: I'm curious if this system actually works outside a perfectly controlled lab setting, Rosa? How robust is this whole setup when things get messy in the real world?

Rosa: That’s a big question, Taro. The paper focuses on how the architecture handles jamming by first having a trajectory generator solve a max-min optimization problem to strengthen that weakest link before the NMPC takes over with its low-level control.

Dev: So this trajectory layer is essentially figuring out where the vehicle needs to go in a way that prioritizes keeping your most important communication link strong, even when there’s interference present. It sets the targets for the NMPC to follow.

Taro: But what happens if the vehicle has to make a sharp maneuver while that optimization is running? Can it keep up with those dynamic changes?

Rosa: The NMPC handles that part, Dev. It takes those high-level references and generates rotor speed commands while making sure the vehicle respects its physical limits and stays dynamically feasible.

Dev: Right, so the NMPC is the inner loop. It’s what actually tells the motors what to do moment by moment, constrained by actuator limits and vehicle dynamics. The paper emphasizes that this NMPC isn't just tracking a path; it’s jointly optimizing motion and communication objectives right there in its cost function.

Taro: That sounds complex because you're mixing physical motion constraints with communication goals simultaneously. How do they manage the trade-off between moving fast and keeping the antenna pointing where it needs to be?

Rosa: They handle that by embedding antenna alignment directly into the NMPC's cost and constraints. Instead of using complicated radiation pattern models, they use geometric alignment metrics to force the antennas into a good position during control.

Title and authors: Dev: That geometric treatment is key because it makes the optimization tractable for real-time computation, which is what you need for an actual flight controller running on limited hardware. They use a mutual directional-alignment constraint, which sets a minimum joint alignment level between the two links.

Taro: So they aren't just hoping the vehicle moves well enough; they are actively forcing the antennas to line up according to some mathematical rule that keeps them balanced across both relay links.

Rosa: That’s right. And to make it work for optimization, they use a second-order Taylor expansion to locally approximate the inverse capacity function, which lets them write a quadratic cost function that's compatible with standard programming methods.

Dev: That approximation is what makes the math work fast enough for the NMPC loop rate, but I always have to remember that it’s only a local approximation, so you can’t expect perfect accuracy everywhere.

Taro: And what about performance compared to simpler methods? The summary mentioned a big jump in capacity—is that improvement realistic when we look at real-world scenarios under heavy jamming?

Rosa: The simulation results show that enforcing those antenna alignment constraints really helps mitigate the deep capacity drops you see with other approaches. They found that on a tilted-propeller platform, the link quality stayed consistently higher than on a coplanar one.

Dev: And for the coplanar vehicle, they saw reduced performance and even temporary connectivity loss near the end of the mission. But here’s something interesting from their tests: under ON-OFF jamming, this controller actually preserves alignment during active interference and brings throughput back right away when the jammer turns off.

Taro: So it shows resilience to intermittent degradation rather than just surviving a single bad patch. That persistence when the interference changes is what really stands out for autonomy.

Rosa: It does. They found that their framework achieves minimum end-to-end capacity numbers approximately ninety-four times larger than Baseline A and fifteen times larger than Baseline B, which shows the performance gain they're aiming for in a practical sense.

Dev: That massive improvement is achieved by continuously redistributing control effort while strictly adhering to actuator limits and keeping that geometric alignment correct. The paper also points out that this approach is agnostic to the specific antenna pattern, as long as it stays doughnut-shaped, which makes it more versatile.

Title and authors: Taro: So for someone interested in building a system that can handle varied hardware or unpredictable interference patterns, this modular structure—the generator feeding the NMPC—seems like a solid starting point.

Rosa: Exactly. It moves away from decoupling motion and communication by making them intrinsically linked throughout the entire control pipeline. But we have to remember what they didn't do: they haven't fully integrated online estimation of jammer characteristics yet, which would be the next logical step for making it even more adaptive.

Dev: Right, and we need to keep an eye on those unmodeled dynamics like aerodynamic coupling and actuator delays when we validate this on real coplanar platforms because the model they used is a simplified GTMR system.

Taro: I agree, the next validation step needs to be testing this framework against those unmodeled dynamics in real flight conditions, not just simulations with perfect models.

Rosa: So we’re seeing a really sophisticated way to tackle jamming by tying the high-level optimization of communication capacity directly into the low-level physical control loop via NMPC and geometric constraints. That's what this paper on Communications Aware NMPC for MultiRotor Aerial Relay Networks Under Jamming Interference is all about.

Dev: It’s a lot of interconnected pieces—from the max-min reference generation to the quadratic reformulation in the cost function—all working together to maintain reliability under adversarial conditions.

Taro: I think what this paper really means for autonomous systems is that we can finally treat communication not as an afterthought but as a primary resource that dictates how we must move and orient our physical bodies.

Rosa: It’s about designing platforms, whether tilted-propeller or coplanar, where the physical geometry itself becomes a key part of the anti-jamming strategy. But remember their limitation: they haven't fully addressed adapting to rapidly varying interference yet.

Dev: So for anyone interested in high-performance relay networks, this paper offers a very concrete framework for joint optimization under jamming constraints. We’ll keep watching how they tackle those next steps in the future work section of Communications Aware NMPC for MultiRotor Aerial Relay Networks Under Jamming Interference.

The paper's summary: Rosa: So, to wrap up what we just talked about, this paper is about designing a smarter way for drone networks to talk when there’s interference, and they achieved a big jump in how much data you can actually move between the drones.

Dev: Right. Basically, they’ve set up a system where the high-level planning figures out the best path by making sure your communication link isn't just about distance but really about keeping that weakest link strong under jamming interference.

Taro: It sounds like they’re trying to solve a problem where you have to balance two completely different things—moving around and staying connected—and the solution is to make those two things work together from the start.

Rosa: Exactly. The core idea is this modular control structure, where you have one part that sets the targets for motion and another part, an NMPC, that handles all the real-time physical movements while constantly checking if those targets are actually achievable with your motors and antennas.

Dev: And they handle the antenna alignment in a really clever way. Instead of using super complicated math based on how radiation patterns work—which is hard to model accurately—they use simple geometric constraints to force the antennas into a good position during maneuvers.

Taro: That’s interesting because it means this whole system can work whether you have a simple coplanar drone or one with tilted propellers, as long as you follow those geometric rules.

Rosa: That’s right. The results show that when they enforce these alignment constraints, the link quality doesn't just stay stable; it stays consistently higher on tilted platforms and even recovers immediately when the jamming signal turns off.

Dev: And the performance numbers are significant. They found that this approach can boost the minimum end-to-end capacity by about ninety-four times compared to their baseline methods, which is a really big jump in terms of network throughput.

Taro: That high number is impressive, but what does it mean for someone just looking at drone autonomy? It means we can actually design aerial relay systems that are much more robust and reliable when they’re operating in challenging environments.

Rosa: It means communication isn't just something you bolt on afterwards; it has to be baked into the way the vehicle plans its whole movement. This framework gives us a way to design platforms where the physical shape of the drone itself helps fight interference.

Dev: The caveat is that they used a local approximation for some of their math, so while it’s great for real-time control, you can't expect absolute perfection in every single tiny calculation.

Taro: And what they didn't cover yet—and what we need to watch—is how this framework handles rapidly changing or adaptive jamming patterns that might move faster than the controller can react.

Rosa: That’s the next big challenge, so we’re looking ahead to future work where they plan to integrate online estimation of the jammer itself so it can adapt its alignment constraints on the fly.

The paper's improvements: Tom: So, we’ve seen how they get those big capacity gains, and now let's talk about what the authors themselves think is missing or what they plan to tackle next with this system.

Rosa: The paper lays out a few clear directions for future research. First, they say that while the current system handles jamming pretty well, it doesn't yet fully account for jammer characteristics that change rapidly or adapt on their own.

Dev: That makes sense from an engineering standpoint because if the interference is moving fast, you need your control loop to be smart enough to adjust its constraints instantly rather than just reacting to a fixed plan.

Taro: I agree, the idea of integrating online estimation of how the jammer is behaving would let the system become truly adaptive, which is crucial if you're dealing with unpredictable adversaries in a real mission.

Rosa: They also flag needing more validation on real-world platforms. They want to test this framework not just in simulation but on actual coplanar and tilted-propeller drones to see how it handles things like aerodynamic coupling or actuator delays that the simplified model might miss.

Dev: Those are practical concerns; if you’re going from a simulation to a real flight controller, those unmodeled dynamics are exactly what can cause trouble. You need to know if the control loop actually stays stable when real-world physics kick in.

Taro: And they mentioned extending the math beyond just Rician fading—the way we model signal loss—to include things like multipath delays or Doppler effects, which are common in complex aerial environments.

Rosa: That moves the channel modeling into a more realistic space, which is important because the current model is simplified for tractability. It shows they are thinking about making it more physically accurate for complex missions.

Dev: From a control perspective, they also point toward embedding an energy-aware term into that NMPC cost function. This would force the system to balance how much power it uses against how much data it’s actually getting, exposing that communication versus energy trade-off clearly.

Taro: That’s interesting because right now the focus is mostly on maximizing capacity under jamming, but adding an energy constraint pushes the system toward making smarter choices about when and where to communicate.

Rosa: It really highlights that the next step isn't just about getting a better number, but about creating a control system that makes trade-offs based on real physical limitations like power consumption.

Dev: So, to summarize, they’re moving from a robust controller for known jamming patterns toward an adaptive one that understands the jammer itself and respects the vehicle's power budget.

Taro: It’s about making the system resilient not just to external noise, but to the changing nature of that noise and your own hardware limitations.

Conclusion: Rosa: So we’ve covered how this paper uses a max-min trajectory generator and an NMPC to jointly optimize motion and communication for Multi-Rotor Aerial Vehicles under jamming interference, and they showed huge gains in capacity.

Dev: Right. We saw the math behind the geometric alignment constraints that replace complicated radiation models to keep the control loop fast enough for real flight hardware.

Taro: It really shows how you can tackle a complex autonomy problem by just tying different levels of control together so they are all working toward one goal, instead of having them fight each other.

Rosa: The implication here is that we can design aerial relay systems where the physical shape of the drone itself becomes a key part of the anti-jamming strategy.

Dev: It means platform design—whether it’s tilted or coplanar—is just as important as writing a clever control algorithm for handling interference.

Taro: And that’s what I think is interesting because it shifts the focus from just making the software smarter to making the physical machine inherently more resilient when things go wrong.

Rosa: Exactly. It moves us toward designing platforms where geometry dictates how well they can communicate, which is a big step for practical applications.

Dev: The main caveat we have is that they haven't fully integrated online jammer estimation yet, so the system relies on a relatively static understanding of the interference environment.

Taro: That’s fair; adapting to truly unpredictable jamming patterns is definitely the next hurdle for this kind of work.

Rosa: It’s a solid framework for what it does now with its modular structure and geometric constraints, especially when looking at the performance numbers they achieved on those tilted platforms versus the coplanar ones.

Dev: We’re ready to look at how this same NMPC approach handles those unmodeled dynamics, like actuator delays and aerodynamic coupling, when we try to test it out on actual flight hardware next.

Taro: I'm curious to see if that framework holds up when you introduce those real-world complexities beyond the ideal mathematical models they used.

Rosa: That’s what we’ll be watching for—the transition from simulation success to real-world reliability for the Communications-Aware NMPC for Multi-Rotor Aerial Relay Networks Under Jamming Interference.

Department of Power Generation Technologies and Materials, Ricerca sul Sistema Energetico S.p.A. · Department of Cybernetics, Czech Technical University in Prague · College of Computing, Mohammed VI Polytechnic University · Department of Engineering, University of Sannio in Benevento · Robotics and Mechatronics Department, University of Twente · Department of Computer, Control and Management Engineering, Sapienza University of Rome

cs.RO

Submitted: 2026-03-30

Updated: 2026-08-19

Comments: This work has been submitted to the IEEE for possible publication

Journal ref: IEEE Transactions on Vehicular Technology, 2026

DOI: 10.1109/TVT.2026.3738530

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 83/100

The gist: The gist: A modular communications-aware control approach combining a high-level max-min trajectory generator with an actuator-level Nonlinear Model Predictive Controller (NMPC) for Multi-Rotor

Key concepts

Max-Min Trajectory Generator
This high-level module solves an optimization problem to find the best path. It specifically targets and strengthens the 'weakest link'—the communication link with the lowest Signal-to-Interference-plus-Noise Ratio. This ensures that overall network capacity is maximized by focusing on the most vulnerable connection.
Nonlinear Model Predictive Controller (NMPC)
This is a low-level controller that manages the vehicle's movements in real time. It tracks the desired path while strictly enforcing physical limits like motor speeds and antenna alignment. It optimizes a cost function that balances tracking accuracy with smooth motion and maintaining optimal antenna orientation.
Geometric Antenna Directionality
Instead of using complex mathematical models of how an antenna radiates signals, this method uses simple geometric constraints to enforce correct alignment. This allows the system to maintain link quality by ensuring the antenna is pointed correctly relative to the vehicle's attitude, regardless of the specific radiation pattern.
Weakest Link Bottleneck
This refers to the single communication link in a network that limits how much data can flow through it. The control strategy focuses on improving this specific bottleneck link first. By optimizing this weakest point, the entire end-to-end capacity of the aerial relay network is substantially increased.

Terminology

Summary

The gist: A modular communications-aware control approach combining a high-level max-min trajectory generator with an actuator-level Nonlinear Model Predictive Controller (NMPC) for Multi-Rotor Aerial Vehicles (MRAVs) under jamming interference achieves a near two-order-of-magnitude increase in minimum end-to-end capacity by jointly optimizing motion and communication objectives while enforcing vehicle dynamics and antenna alignment constraints.

The Problem Context

Multi-Rotor Aerial Vehicles (MRAVs) are increasingly used in communication-dependent missions where connectivity loss directly compromises task execution. Existing antijamming strategies often decouple motion from communication, overlooking that link quality depends on vehicle attitude and antenna orientation. In coplanar platforms, “tilt-to-translate” maneuvers can inadvertently align antenna nulls with communication partners, causing severe degradation under interference. The field of communications-aware robotics addresses these challenges by jointly embedding communication objectives—mobility, connectivity, sensing, and energy—into motion planning and control.

The Proposed Architecture

The paper introduces a modular communications-aware control architecture for MRAVs composed of a trajectory generator and a Nonlinear Model Predictive Controller (NMPC). The trajectory generator solves a max–min optimization problem to strengthen the weakest communication link under jamming. The NMPC then tracks these references while enforcing vehicle dynamics, actuator limits, and antenna-alignment constraints. A key feature is that antenna directionality is incorporated via geometric alignment constraints rather than explicit radiation pattern models.

Trajectory Generation Details

The trajectory generator computes high-level motion references by solving a max–min optimization problem that strengthens the weakest communication link under jamming. To address non-convexity, the method employs a h-norm approximation, second-order Taylor expansions, and a quadratic reformulation. Bandwidth allocation is selected to balance the link capacities and maximize end-to-end throughput. The approach consists of two integrated modules: a high-level trajectory generator that solves a max–min problem to strengthen the weakest RF link under interference, and a low-level NMPC that tracks the references while enforcing vehicle dynamics and actuator constraints.

The NMPC Control Loop

The communications-aware NMPC serves as the inner-loop controller, generating rotor-speed commands while enforcing actuator constraints and dynamic feasibility. Unlike conventional approaches, it jointly optimizes motion and communication objectives by embedding antenna alignment directly into the cost and constraints. A distinctive feature is its geometric treatment of antenna directionality, which avoids explicit radiation-pattern models and preserves alignment during agile maneuvers. The objective function minimizes three objectives—reference tracking, smooth actuation, and antenna alignment—encoded in stage and terminal costs.

Key Constraints and Approximations

The NMPC enforces directional alignment through geometric misalignment metrics and exploits motion intent shared via OCC for coordination. The mutual directional-alignment constraint is (1−v1,2)(1−v1,0) ≥ µ, which sets the minimum joint alignment level. To obtain a quadratic-programming-compatible cost function for real-time optimization, the inverse capacity function is locally approximated by a second-order Taylor expansion. This reformulation replaces explicit radiation-pattern models with geometric alignment constraints, yielding a tractable optimization that preserves the dominant dependence on range, antenna orientation (through p¨ d,1), and jammer interference.

Simulation Results and Findings

Simulations show that enforcing antenna-alignment constraints mitigates deep capacity drops and stabilizes throughput. The tilted-propeller platform maintains consistently higher link quality, whereas the coplanar vehicle exhibits reduced performance and a temporary connectivity loss near the end of the mission. Under ON–OFF jamming, the controller preserves alignment during active interference and restores throughput immediately when the jammer turns OFF, confirming resilience to intermittent degradation. The proposed framework achieves Cmin approximately 94× larger than Baseline A and 15× larger than Baseline B.

Practical Implications

The results have three practical implications for aerial relay system design: First, the directional coupling between attitude and link quality is non-negligible whenever the relay carries non-isotropic antennas. Second, actuation geometry is a first-class design parameter for jamming resilience: tilted-propeller (or fully-actuated) platforms preserve directional alignment under relay maneuvers, while coplanar designs trade directional fidelity for mechanical simplicity. Third, the framework’s modular structure—max–min reference generator coupled with a geometric-alignment NMPC—is agnostic to the specific antenna pattern, provided it remains doughnut-shaped.

Future Work

Future work will focus on four directions: First, integrating online estimation of jammer characteristics and adaptive alignment constraints for resilience to rapidly varying or adaptive interference. Second, validating the framework on real coplanar and tilted-propeller MRAV platforms to assess unmodeled dynamics such as aerodynamic coupling, actuator delays, and sensor latency. Third, extending the channel model beyond Rician fading (Section VII-F) to incorporate shadowing, multipath delay spread, and Doppler effects. Fourth, embedding an energy-aware term into the NMPC cost function as a cost term or energy constraint to expose the communication-energy trade-off.

How it works

The system integrates two complementary communication technologies: a RF link for high-throughput data transmission and an Optical Camera Communication (OCC) channel for inter-agent coordination. The RF link enables MRAV-2 to transmit data to MRAV-1, which relays it to the BS. Coordination under jamming conditions is supported by an OCC system, such as the UltraViolet Direction and Ranging platform which provides a jamming-immune, lowlatency channel for MRAV-to-MRAV communication.

System Dynamics and Constraints

The MRAV is modeled as a GTMR system actuated by Np motor-propeller units arbitrarily placed and oriented with respect to the body. The state x comprises position p ∈ R cubed and linear velocity v = p˙ ∈ R cubed (both in FW), orientation η = (ϕ,ϑ,ψ)⊤ ∈ R cubed (Euler angles), and angular velocity ω ∈ R cubed (in FB). The augmented state is x¯ = (p⊤,η⊤,v⊤,ω⊤,u⊤)⊤ ∈ R(12+Np). This formulation enables explicit actuator constraints: γ ≤ u ≤ γ¯ and γ˙ ≤ u˙ ≤ γ˙¯.

Communication Channel Modeling

The RF links are modeled using the free-space path loss model under unobstructed LoS conditions. Each MRAV carries a halfwave dipole antenna rigidly mounted along its body-frame zBaxis, coupling the antenna’s orientation in FW directly to vehicle attitude η and making link performance sensitive to motion-induced misalignment.

Improvements for AI systems

  1. Bold trajectory generator with max-min optimization: The trajectory generator solves a max–min optimization problem to strengthen the weakest communication link under jamming, producing communications-aware motion references that feed the NMPC as a timevarying reference rather than an open-loop plan. This allows the system to prioritize link quality over simple distance minimization.

  2. Bold mutual directional-alignment constraint: The NMPC enforces a mutual directional-alignment constraint (20)—jointly coupling both relay links through a product-form inequality with slack-based feasibility recovery, which enforces balanced antenna orientation across both links, consistent with end-to-end performance being bottlenecked by the weakest one. This prevents link degradation from antenna misalignment.

  3. Bold geometric treatment of antenna directionality: The method handles antenna directionality through geometric alignment constraints rather than explicit radiation-pattern parametrization, which avoids explicit radiation-pattern models and preserves alignment during agile maneuvers. This makes the control architecture compatible with a wide range of multirotor designs, from coplanar to tilted-propeller platforms.

  4. Bold unified evaluation across configurations: The framework provides a unified evaluation, on a common Generically-Tilted Multi-Rotor (GTMR) dynamic model, of coplanar and tilted-propeller actuation, allowing researchers to isolate the influence of actuation geometry—as opposed to algorithmic choices—as p1 p2. This prevents misinterpreting performance differences as solely due to control strategy.

  5. Bold robustness under time-varying jamming: The controller preserves alignment during active interference and restores throughput immediately when the jammer turns OFF, confirming resilience to intermittent degradation, which is validated by simulation showing that the ON–OFF pattern does not introduce additional outages. This means the system maintains high performance even when adversaries switch jamming patterns.

  6. Bold comparison with external baselines: The proposed framework achieves Cmin approximately 94× larger than Baseline A and 15× larger than Baseline B, demonstrating that the approach overcomes the limitations of traditional methods by continuously redistributing control effort while strictly adhering to actuator limits and preserving geometric alignment.

Abstract

Multi-Rotor Aerial Vehicles (MRAVs) are increasingly used in communication-dependent missions where connectivity loss directly compromises task execution. Existing anti-jamming strategies often decouple motion from communication, overlooking that link quality depends on vehicle attitude and antenna orientation. In coplanar platforms, ``tilt-to-translate'' maneuvers can inadvertently align antenna nulls with communication partners, causing severe degradation under interference. This paper presents a modular communications-aware control approach combining a high-level - trajectory generator with an actuator-level Nonlinear Model Predictive Controller (NMPC). The trajectory layer optimizes the weakest link---the link with the lower instantaneous Signal-to-Interference-plus-Noise Ratio, which bottlenecks the harmonic-mean end-to-end capacity---while the NMPC enforces vehicle dynamics, actuator limits, and antenna-alignment constraints. Antenna directionality is handled geometrically, avoiding explicit radiation-pattern parametrization. The method is evaluated in a relay scenario with an active jammer and compared across coplanar and tilted-propeller architectures. Results show a near two-order-of-magnitude increase in minimum end-to-end capacity, markedly reducing outage events, with moderate average-capacity gains. Tilted platforms preserve feasibility and link quality, whereas coplanar vehicles show recurrent degradation. These findings indicate that full actuation is a key enabler of reliable communications-aware operation under adversarial directional constraints.

Related papers