SEP-NMPC: Safety Enhanced Passivity-Based Nonlinear Model Predictive Control for a UAV Slung Payload System
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.
Dev: And I'm Dev, here with Taro.
Rosa: We're looking at the paper: "SEP-NMPC: Safety Enhanced Passivity-Based Nonlinear Model Predictive Control for a UAV Slung Payload System".
Title and authors: Rosa: So we're looking at this paper called "SEP-NMPC: Safety Enhanced Passivity-Based Nonlinear Model Predictive Control for a UAV Slung Payload System" and it’s about making sure a drone carrying something slung underneath it doesn't crash.
Dev: Right, so the core idea here is using Model Predictive Control, or MPC, which is great for trajectory optimization online with all those constraints on inputs and states. But as you know, when you have a payload suspended like that under a quadrotor, keeping things stable while avoiding obstacles gets really tricky.
Taro: I mean, the authors are aiming to give formal guarantees of stability and safety in cluttered environments, which is tough because the dynamics are so coupled—the drone and the payload move together.
Rosa: Exactly. They introduce this Safety Enhanced Passivity-Based Nonlinear MPC, or SEP-NMPC, which they claim provides those formal guarantees for a quadrotor transporting a slung payload through cluttered environments.
Dev: The stability part comes from embedding a strict passivity inequality right into the optimization problem itself, which they say is derived from a shaped energy storage function with adaptive damping.
Taro: That sounds like it’s trying to actively dissipate excess energy and ensure asymptotic convergence even when the payload is swinging around.
Rosa: And for safety, they’re using high-order control barrier functions, or HOCBFs, which they state render user-defined clearance sets forward invariant.
Dev: So those HOCBFs are what obliges both the quadrotor and the suspended payload to maintain a certain separation while dealing with static and dynamic obstacles.
Taro: That’s important because it means we’re not just hoping for avoidance; we have mathematical proof that they will stay out of danger zones.
Rosa: They also mention that this whole optimization framework remains quadratic-program compatible, which is a big deal for real-time implementation on onboard processors.
Dev: And they show how the passivity constraint couples the position error with the payload swing energy in a shaped energy function, which captures that underactuated structure between the vehicle and the load.
Taro: So it’s not just controlling one thing; it’s modeling how that swinging load affects everything else, which is a lot of coupled dynamics to handle.
Rosa: And they also discuss how this unified framework allows them to enforce both system stability and safety constraints seamlessly within a single optimization problem at each sampling instant.
Dev: That means no switching between different safety layers or heuristic decisions when things get complicated; it runs online without gain scheduling or that kind of guesswork.
Taro: When the world misbehaves—say, an obstacle suddenly appears—how does this system handle the sudden change in dynamics and still recover gracefully?
Rosa: That’s where the HOCBFs come in; they are constructed by stacking conditions for each obstacle–body pair, resulting in constraints like attitude controller reference trajectory plus motion capture obstacles/environment data.
Dev: And those constraints, which involve affine inequalities, are what guarantee forward invariance of safety sets while staying computationally tractable for real-time use on lightweight onboard processors.
Taro: So it keeps the complexity low enough to run fast while still being rigorous about keeping things away from walls or other objects.
Rosa: It seems like they’ve managed to unify stability, which is the passivity part, with safety through those barrier functions in one single optimization problem for this UAV slung payload system.
Dev: The stability part specifically uses that shaped energy function with adaptive damping to ensure asymptotic stability without needing manual gain tuning.
Taro: That’s a nice touch; avoiding the need for someone to fine-tune parameters just to get the fundamental stability working.
Rosa: So, what does this mean for people who only listen to this show? It means we're talking about a system that can navigate cluttered areas reliably, not just in ideal lab settings but in real-world scenarios with moving things around it.
Dev: And the results on the Quanser QDrone2 carrying a cable–suspended point-mass payload at one hundred hertz sampling rate with a two second prediction horizon show they are actually achieving safe obstacle avoidance and stable trajectory tracking consistently.
Taro: The experimental validation seems to confirm that adding passivity improves smoothness and safety, but this SEP-NMPC is what achieves the most reliable avoidance and convergence when things get tricky.
Rosa: So, to wrap up on the technical side, they’ve successfully introduced a safety enhanced passivity-based nonlinear MPC for quadrotor–payload transport that unifies passivity–based stability with high-order CBF safety.
Dev: It keeps the optimization quadratic-program compatible and runs online without gain scheduling or heuristic switching, offering reliable performance even in cluttered, dynamic settings.
Taro: If I had to weigh in on what this means for the future, it suggests a path toward more robust real-time control for systems with complex interactions like drones carrying payloads.
Rosa: It really does suggest that we can build controllers that handle both the inherent dynamics of the system and the physical limitations of operating in a real, messy environment simultaneously.
Dev: The main thing is this unified framework allows them to enforce both stability and safety constraints in one go, which is a significant simplification for designing real-time systems.
Taro: It moves us closer to systems that can operate reliably when things aren't perfectly predictable because they have these formal guarantees built in.
Rosa: That’s the gist of what they presented with this SEP-NMPC work. We’ll be looking at more papers on how other systems handle uncertainty next time.
The paper's summary: Rosa: So, to recap, this paper is about taking Model Predictive Control for drones carrying things slung underneath them and making it safer by adding two main things: passivity theory for stability and control barrier functions for collision avoidance.
Dev: Right, that’s the gist of SEP-NMPC. It solves a big optimization problem at every moment to find the best path while making sure the system stays stable and doesn't hit anything.
Rosa: The paper claims this approach gives formal guarantees on both stability and safety in cluttered areas without needing manual tuning or guesswork when things get messy.
Dev: They embed a strict passivity inequality into that optimization problem, which they use to dissipate excess energy and keep the system converging asymptotically, even with the payload swinging around.
Rosa: And on top of that, they stack high-order control barrier functions for safety, which means every obstacle or body pair has its own constraint to keep them separated.
Dev: Those barrier functions are constructed so they guarantee the clearance sets remain forward invariant, meaning you can trust that the system stays out of danger zones when interacting with moving obstacles too.
Rosa: The authors show it works on a real drone setup, running in real time at one hundred hertz sampling rate, and they confirm it keeps tracking the path and avoiding obstacles consistently.
Dev: It's interesting how they manage to unify those two concepts—the stability from passivity and the safety from barrier functions—into one optimization framework that runs online without needing any heuristic switching or gain scheduling.
Rosa: That unification is what really stands out for me, because usually you’d have to choose between a controller that's stable but maybe too slow, or one that’s safe but computationally too heavy for a drone.
Dev: Exactly. The paper shows this SEP-NMPC remains quadratic program compatible, which means it can actually run on the onboard processors in real time without huge latency issues.
Rosa: So what does this mean for us out here, where we're testing these things outside the lab? It suggests that we can build more robust systems for complex real-world tasks, not just nice simulations.
Dev: It moves us closer to having drone systems that can handle dynamic interactions from moving intruders while still maintaining a predictable and stable trajectory.
Rosa: The next thing I want to look at is how they specifically modeled the coupled dynamics between the vehicle and that swinging load, because modeling those two moving parts together is always the hardest part of this kind of control.
The paper's improvements: Taro: So, we just talked about how SEP-NMPC uses passivity and barrier functions to combine stability and safety in one optimization problem for quadrotors carrying payloads.
Rosa: Right, but this paper also points out some specific improvements they made to that core idea that are pretty significant for real-world use.
Dev: They focus on how they shaped the energy storage function used for passivity, making it specifically couple the position error with the payload swing energy in a way that really captures those underactuated dynamics.
Rosa: It’s like they didn't just add a random damping term; they made it smarter so that when the payload swings, you get this adaptive damping that actually helps stabilize everything without needing to manually tune parameters.
Dev: That’s key because manual tuning is a headache and often fails when things change in the environment, so having that adaptive mechanism built-in is a big step for robustness.
Taro: And then they talk about how the control barrier functions are constructed; instead of just checking if you're clear of an obstacle at one point, they stack those conditions for every possible obstacle–body pair.
Rosa: That means the safety check is really comprehensive; it’s not just about the nearest wall, but accounting for all these dynamic interactions between the drone and everything else around it.
Dev: From a controls engineer's view, that stacking of affine inequalities is what allows them to keep the solver computationally tractable while still enforcing those hard safety boundaries in real time.
Taro: It means when things get chaotic, like an unexpected gust or a sudden movement from another object, the system has this layered defense mechanism built in by those barrier functions.
Rosa: So what does this mean for someone who just watches the drone on their phone? It means the drone is designed to handle a lot of messy, unpredictable situations safely, and it's doing that in real time.
Dev: The main implication is that we can move toward systems where safety isn't just an afterthought or something you patch in later; it’s baked into the fundamental control law from the start.
Taro: It suggests a path toward autonomy where the system doesn't just follow a path, but actively manages its stability and its boundaries against a constantly changing physical world.
Rosa: I think what’s really cool is how they keep the optimization framework compatible with quadratic programs, which keeps it fast even when those complex safety constraints are being evaluated.
Dev: Yes, that QP compatibility is what makes it practical for deployment on actual hardware instead of just theoretical papers and simulations.
Taro: So for future work, I'm thinking they should explore how this framework handles situations where the environment itself is changing in a very unpredictable way, maybe something outside the scope of their current fixed clearance sets.
Conclusion: Rosa: So to wrap up, this SEP-NMPC paper is about taking Model Predictive Control for quadrotors carrying payloads and making it safer by combining passivity stability with control barrier functions for obstacle avoidance.
Dev: It gives formal guarantees on both stability and safety in cluttered areas using a unified optimization framework that runs in real time without needing manual tuning.
Rosa: The implication is that we can build drone systems that handle complex, messy real-world scenarios reliably, not just in controlled lab settings.
Dev: Right, it’s about moving toward control systems where safety isn't an afterthought; it’s built into the fundamental law of the controller itself.
Taro: I think what they show with this SEP-NMPC is how to create autonomy that can gracefully recover when things go wrong in dynamic environments.
Rosa: It really shows that combining stability and safety constraints in one optimization problem makes real-time control much more practical for something like a drone.
Dev: The paper proves the concept works on a Quanser QDrone2, which gives us concrete numbers showing it achieves safe obstacle avoidance and stable tracking at one hundred hertz sampling rate.
Taro: That real-world validation is important because it shows the theory holds up when you actually try to deploy it outside of just writing equations.
Rosa: It’s a solid piece of work, this SEP-NMPC paper, because it gives us a very robust way to design controllers for coupled systems.
Dev: The next thing we should look at is how other researchers are tackling the computational cost of these kinds of complex models when they try to run them on smaller embedded hardware.
Seyedreza Rezaei, Junjie Kang, Amaldev Haridevan, Jinjun Shan
York University
cs.RO, cs.SY, eess.SY
Submitted: 2026-03-09
Updated: 2026-03-09
Comments: Accepted at ICRA 2026
Journal ref: 2026 IEEE International Conference on Robotics and Automation (ICRA), Vienna, Austria, 2026, pp. 16007-16014
DOI: 10.1109/ICRA57385.2026.11696829
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 83/100
The gist: Model Predictive Control (MPC) is widely adopted for UAV slung payload systems as it enables online trajectory optimization with input and state constraints.
Key concepts
- Model Predictive Control (MPC)
- MPC is an optimization technique used for controlling systems, like a drone. It looks ahead over a short time horizon, calculates the best control inputs to minimize a cost function while respecting physical limits (constraints), and repeats this process in real-time. This allows it to plan trajectories that account for future obstacles.
- Passivity-Based Stability
- This method enforces stability by embedding a mathematical condition called passivity directly into the optimization problem. It uses a shaped energy storage function that adapts damping based on system state, ensuring the system naturally dissipates excess energy and converges to a stable state without needing manual gain tuning.
- High-Order Control Barrier Functions (HOCBFs)
- HOCBFs are mathematical functions used to define safe regions around obstacles. By constraining the control inputs based on these functions, they ensure that the system's state remains within a predefined 'safe set,' guaranteeing collision avoidance even when dealing with complex, swinging payloads.
Terminology
Summary
Model Predictive Control (MPC) is widely adopted for UAV slung payload systems as it enables online trajectory optimization with input and state constraints. p. 0 This paper introduces a Safety Enhanced Passivity-Based Nonlinear MPC (SEP-NMPC) that provides formal guarantees of stability and safety for a quadrotor transporting a slung payload through cluttered environments. p. 0
System Dynamics Description
The system dynamics are derived using the Lagrangian formulation, resulting in a cascaded underactuated structure with eight degrees of freedom (DOFs) and only four control inputs. The translational dynamics are expressed as M(q)¨q + C(q, q˙) ˙q + G(q) = u. p. 1 The inertia matrix M is given by M = (mQ + mL)I3 Mc / Mc. p. 1 The rotational dynamics are described by R˙ = Rω× and Jω˙ = τ − ω ×(Jω). p. 1
SEP-NMPC Formulation
The SEP-NMPC solves an optimization problem at each sampling instant tk, minimizing the cost function J = Z tk+Ttk / dt (7a) s.t. x˙(t) = f x(t), ua(t), x(tk) = xk, (7b) and other constraints. p. 0 The optimization includes a strict passivity condition where ua(t)⊤v(t) ≤ − ρ∥v(t)∥2 − ε∥ua(t)∥2, (7d). p. 3 Safety is guaranteed through high-order control barrier functions (HOCBFs) that render user-defined clearance sets forwardinvariant. p. 0
Passivity-Based Stability
The stability is enforced by embedding a strict passivity inequality, which is derived from a shaped energy storage function with adaptive damping, directly into the NMPC. The enhanced storage function is defined as V (x) = 1/2 q˙⊤M(q) ˙q + mLgl1 − cos α cos β + 1/2 e⊤ ζ Keζ. p. 2 This formulation ensures asymptotic stability without resorting to gain tuning. p. 0
Safety Guarantees via HOCBFs
Safety is guaranteed through high-order control barrier functions (HOCBFs) that explicitly account for both the quadrotor and the suspended payload. The constraints are constructed by stacking these conditions for each obstacle–body pair, resulting in ACBFx(t), t ua(t) ≥ bCBFx(t), t. p. 2 This ensures forward invariance of the safety sets while remaining computationally tractable for real-time implementation. p. 1
Experimental Validation
The SEP-NMPC was implemented on a Quanser QDrone2 carrying a cable–suspended point-mass payload, running in real time at 100 Hz with a prediction horizon of T = 2 s. The results confirm that SEP-NMPC consistently achieves safe obstacle avoidance and stable trajectory tracking in real time. p. 5 Adding passivity improves smoothness and safety, while SEP-NMPC achieves the most reliable avoidance and convergence. p. 6
Conclusion
This work introduced a Safety Enhanced Passivity-Based NMPC for quadrotor–payload transport that unifies passivity–based stability with high-order CBF safety. The resulting optimization remains QP-compatible and runs online without gain scheduling or heuristic switching, offering reliable performance even in cluttered, dynamic settings. p. 7 The combined results verify that SEP-NMPC consistently achieves safe obstacle avoidance and stable trajectory tracking in real time. p. 5
REFERENCES
[1] A. Ollero, G. Heredia, A. p. 7 Franchi, G. p. 7 Antonelli, K. p. 7 Kondak, A. Sanfeliu, A. p. 7 Viguria, J. p. 7 R. Martinez-de Dios, F. p. 7 Pierri, J. Cortes, A. Santamaria-Navarro, M. p. 7 A Trujillo Soto, R Balachandran, J Andrade-Cetto and A Rodriguez. p. 7
[2] T Tomic, K Schmid, P Lutz, A Domel, M Kassecker E Mair I L Grixa F Ruess M Suppa and D Burschka. p. 7
[3] I Palunko P Cruz and R Fierro. p. 7
[4] A Ollero M Tognon A Suarez D Lee and A Franchi. p. 7
[5] T Lee K Sreenath and V Kumar. p. 7
[6] S Tang and V Kumar. p. 7
[7] G Tartaglione E D’Amato M Ariola P S Rossi and T A Johansen. p. 7
[8] S Sun A Romero P Foehn E Kaufmann and D Scaramuzza. p. 7
[9] P Foehn D Falanga N Kuppuswamy R Tedrake and D Scaramuzza. p. 7
[10] W Xiao and C Belta. p. 7
[11] X Tan W S Cortez and D V Dimarogonas. p. 7
[12] A Pallar G Li M Sarvaiya and G Loianno. p. 7
[13] L F Recalde M Sarvaiya G Loianno and G Li. p. 7
[14] T K Tasooji S Khodadadi and G Liu. p. 7
[15] G Li and G Loianno. p. 7
[16] X Liang Y Fang N Sun and H Lin. p. 7
[17] J Kang J Shan and H Alkomy. p. 7
[18] F Kendoul. p. 7
[19] B Xian S Wang and S Yang. p. 7
[20] T Raff C Ebenbauer and P Allgower. p. 7
[21] J Kang and Z H Zhu. p. 7
[22] A D Ames X Xu J W Grizzle and P Tabuada. p. 7
[23] J Zeng B Zhang and K Sreenath. p. 7
[24] R Verschueren G Frison D Kouzoupis J Frey N van Duijkeren A Zanelli B Novoselnik T Albin R Quirynen and M Diehl. p. 7
[25] J A E Andersson J Gillis G Horn J B Rawlings and M Diehl. p. 7
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
- Stability Enhancement through Passivity-Based Control
The system can achieve asymptotic stability of an underactuated quadrotor–slung payload system by embedding a strict passivity inequality into the Model Predictive Control optimization, which dissipates excess energy and ensures asymptotic convergence despite payload swings.
- Formal Safety Guarantees for Obstacle Avoidance
The improved AI system can guarantee collision-free trajectories by using high-order control barrier functions (HOCBFs) that render user-defined clearance sets forward invariant, obliging both the quadrotor and the swinging payload to maintain separation while interacting with static and dynamic obstacles.
- Unified Real-Time Optimization Framework
The SEP-NMPC framework allows for a single optimization problem that enforces both system stability and safety constraints seamlessly, without heuristic switching, and runs in real time,
solving it online at each sampling time.
- Coupled Dynamics Modeling
The system can accurately model the coupled dynamics between the vehicle and the swinging load by defining a shaped energy function that couples position error with payload swing energy
to capture the system's underactuated structure.
- Robust Performance in Cluttered Environments
The resulting controller ensures collision–free navigation, smooth convergence, and consistently low solver times,
maintaining performance even when facing dynamic interactions from moving intruders.
Sources
- ES-HPC-MPC: Exponentially Stable Hybrid Perception Constrained MPC for Quadrotor with Suspended Payloads
- Event-Triggered Nonlinear Model Predictive Control for Cooperative Cable-Suspended Payload Transportation with Multi-Quadrotors
Related papers
- Rethinking Actions: Continuity-Aware Priors for World-Model Planning
- MOSAIC: Modular Obstacle and Spatial-Aware Framework for Manipulation In Cluttered Scenes
- SAGAS: Semantic-Aware Graph-Assisted Stitching for Offline Temporal Logic Planning
- DexSeed: Expanding Human Hand-Object Interactions into Robot Demonstrations for Dexterous Policy Learning
- LaST0.5: A Generalist World Action Model with Interaction-Centric Latent Reasoning
- CIALLO: Learning Shared Intent Representations for Cross-Embodiment Manipulation