RCOA Extension and Applications

arXiv:2607.02797 · eess.SY, cs.SY · Submitted 2026-07-02 · 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: We're looking at the paper: "RCOA Extension and Applications".

Dev: Here's the gist of the paper. The Relaxed Convex Obstacle Avoidance (RCOA) formulation is presented as a first approach to enable a fully convex optimal control problem (OCP) for obstacle avoidance.

Rosa: This framework is extended to three-dimensional environments and applied to Unmanned Aerial Vehicle (UAV) navigation.

Dev: Core Contribution The RCOA formulation yields an analytical framework that defines a unique characteristic: the ability to maintain obstacle avoidance (OA) efficacy even when obstacles reside beyond the controller’s prediction horizon.

Rosa: This extension incorporates vehicle geometries, moving beyond point-mass representations to enable collision avoidance between 3D objects.

Title and authors: Rosa: So we're looking at this paper now called "RCOA Extension and Applications". It sounds like they are taking an existing idea about obstacle avoidance, RCOA, and extending it for three dimensions to use with unmanned aerial vehicles.

Dev: Yeah, the title suggests they’re showing how to take that original Relaxed Convex Obstacle Avoidance formulation and apply it to a real problem where things are in three dimensions.

Rosa: What does that mean for us on the ground? It means we're not just talking about simple 2D avoidance anymore; we're talking about navigating space like a drone, which is way more complex because there are height and depth constraints too <ref:2607.02797#pg2>.

Dev: Right. And the authors, Ricardo Tapia and Iman Soltani, they’ve got a background in optimization and control theory which tells you this isn't just some casual thought; it's built on a solid mathematical foundation for real-time systems.

Rosa: They are taking something that was originally designed to make an optimal control problem fully convex so you can solve it easily, and they’re proving that this framework still works when you add the complexity of three dee geometry <ref:2607.02797#pg1>.

Dev: The implication is they're tackling a problem where standard methods get too computationally expensive for the fast loops we need in autonomous flight. They’re aiming for something that runs quickly enough to actually fly safely.

The paper's summary: Rosa: So, this paper summarizes how RCOA allows you to keep the obstacle avoidance working even if obstacles are way outside what your controller can currently predict, which is a huge deal.

Dev: That’s the horizon independence thing they highlight; it means the analytical framework gives you guidance even when things are beyond your immediate look-ahead window, which is super useful when you have to make quick evasive moves.

Rosa: They also mention they moved beyond just treating obstacles as little dots, using vehicle geometries instead, which lets them check for collisions between full three dee objects without being overly cautious about the space around them <ref:2607.02797#pg1>.

Dev: And they found that this approach doesn't just work mathematically; it actually gives you computational performance on par with or even better than other methods we’ve seen before when solving the optimal control problem for quadrotor dynamics.

Rosa: So, to put it simply, they’ve taken a method that handles 2D avoidance and shown that by adding vehicle shape and going to three dee, you still get a fast solution that's robust enough for drone navigation <ref:2607.02797#pg2>.

Dev: The key summary point is that this formulation lets the nonlinear model predictive controller execute aggressive maneuvers through tight spots even when you are using a shorter prediction horizon than usual, keeping those control frequencies above thirty hertz.

The paper's improvements: Rosa: One major improvement they point out is how they handle the scaling issues that come with navigating in three dimensions, which is a big hurdle for aerial navigation systems.

Dev: They address that by specifically looking at how to transition from just using a single point mass representation of the vehicle to using a multi-point approximation of the actual vehicle geometry.

Rosa: That means they’re getting better fidelity when checking collisions because they aren't just guessing where the center of the drone is; they are modeling its shape properly.

Dev: And that leads to another improvement, which is what they call horizon independence, where the analytical formulation provides a guidance effect for obstacles residing outside the prediction horizon.

Rosa: That means even if you’re in a tricky situation and you only have a small time window to plan, this method still gives you a safe direction based on the continuous penalty gradient.

Dev: And from an engineering standpoint, they show that this formulation is computationally superior to other obstacle avoidance formulations like point-to-ellipsoid or dual minimum distance methods because it’s consistent and fast in the real-time loop.

Conclusion: Rosa: So, to wrap up on "RCOA Extension and Applications," they’ve shown that this formulation is a strong way to solve obstacle avoidance for drones by handling three dee space and geometry without sacrificing the speed needed for real-time control <ref:2607.02797#pg1>.

Dev: The main point is that because of the continuous penalty gradient derived from KKT conditions, the nonlinear model predictive controller can inherently anticipate obstacles and execute evasive maneuvers independently of how long its prediction horizon is set.

Rosa: It’s about making the controller proactive rather than just reactive to what's right in front of it, which is a significant shift for autonomous systems operating in complex environments.

Dev: The paper shows that this method provides computational consistency that’s much tighter than other methods we tested, which means you can rely on the solver running predictably under stress.

Taro: From an autonomy perspective, the fact that it handles obstacles outside the prediction horizon is key because in real life, things don't always behave exactly as modeled; this formulation seems to give the system a way to stay safe even when things get messy.

Rosa: Exactly. It’s not just about solving a math problem fast; it’s about giving the drone intelligence to maneuver safely through tight spots, even when its planning window is short.

Dev: So, this RCOA Extension and Applications paper gives us a concrete way to build more reliable flight software that can handle the geometry and speed requirements of modern UAVs while keeping the control loop running at high frequencies.

University of California at Davis

eess.SY, cs.SY

Submitted: 2026-07-02

Updated: 2026-11-13

Comments: 11 pages, 9 figures, multimedia

License: http://creativecommons.org/licenses/by-nc-sa/4.0/

Importance score: 89/100

The gist: The Relaxed Convex Obstacle Avoidance (RCOA) formulation is presented as a first approach to enable a fully convex optimal control problem (OCP) for obstacle avoidance.

Key concepts

Relaxed Convex Obstacle Avoidance (RCOA)
This formulation is a mathematical approach that simplifies a difficult optimal control problem into one that can be solved efficiently. Its key feature is that it provides an analytical framework for obstacle avoidance, allowing the controller to anticipate obstacles beyond its short-term prediction window, ensuring safety even in challenging situations.
Vehicle Geometry Approximation
Instead of treating the vehicle as a single point, this method uses a multi-point approximation of the vehicle's shape. This allows for more accurate collision checking between 3D objects while still keeping the calculations manageable and avoiding overly conservative avoidance strategies that might prevent necessary maneuvers.
Horizon Independence
This is RCOA's unique property where the mathematical formulation provides a guidance effect for obstacles located outside the controller's prediction horizon. This means the system can safely plan evasive actions based on anticipated obstacle locations, even if those obstacles are not within the immediate look-ahead window.
KKT Derived Continuous Penalty Gradient
This is the mathematical mechanism that allows RCOA to work. It generates a continuous penalty gradient derived from Karush-Kuhn-Tucker (KKT) conditions. This gradient acts as an anticipatory signal, guiding the controller toward safe paths well before a hard boundary is reached, enabling proactive obstacle avoidance.

Terminology

Summary

The Relaxed Convex Obstacle Avoidance (RCOA) formulation is presented as a first approach to enable a fully convex optimal control problem (OCP) for obstacle avoidance. p. 0

Where the paper says it

1 RCOA Extension and Applications Ricardo Tapia, Iman Soltani Abstract—The Relaxed Convex Obstacle Avoidance (RCOA) formulation is the first approach to enable a fully convex optimal control problem (OCP) for obstacle avoidance. Convergence analysis of RCOA yields an analytical framework that defines a unique characteristic: the ability to maintain obstacle avoidance (OA) efficacy even when obstacles reside beyond the controller’s prediction horizon.

This framework is extended to three-dimensional environments and applied to Unmanned Aerial Vehicle (UAV) navigation. p. 0
Where the paper says it

Convergence analysis of RCOA yields an analytical framework that defines a unique characteristic: the ability to maintain obstacle avoidance (OA) efficacy even when obstacles reside beyond the controller’s prediction horizon. In this paper, RCOA is extended to three-dimensional environments and apply it to Unmanned Aerial Vehicle (UAV) navigation.

Core Contribution

The RCOA formulation yields an analytical framework that defines a unique characteristic: the ability to maintain obstacle avoidance (OA) efficacy even when obstacles reside beyond the controller’s prediction horizon. This extension incorporates vehicle geometries, moving beyond point-mass representations to enable collision avoidance between 3D objects. p. 0

Where the paper says it

In this paper, RCOA is extended to three-dimensional environments and apply it to Unmanned Aerial Vehicle (UAV) navigation. Furthermore, the formulation is enhanced to incorporate vehicle geometries, moving beyond point-mass representations to enable collision avoidance between 3D objects.

Numerical simulations show that RCOA provides computational performance on par or exceeding state-of-the-art methods. p. 0
Where the paper says it

Furthermore, the formulation is enhanced to incorporate vehicle geometries, moving beyond point-mass representations to enable collision avoidance between 3D objects. Numerical simulations demonstrate that RCOA provides computational performance on par or exceeding stateof-the-art methods.

Furthermore, it enables a Nonlinear Model Predictive Controller (NMPC) to execute aggressive maneuvers through narrow passages with reduced prediction horizons, ensuring real-time feasibility at frequencies exceeding 30 Hz. p. 0
Where the paper says it

Numerical simulations demonstrate that RCOA provides computational performance on par or exceeding stateof-the-art methods. Notably, RCOA is demonstrated to enable a Nonlinear Model Predictive Controller (NMPC) to execute aggressive maneuvers through narrow passages with reduced prediction horizons, ensuring real-time feasibility at frequencies exceeding 30 Hz.

Methodological Extensions

The paper extends RCOA in three critical dimensions. These extensions include:

  1. 3D Scaling: Extending the RCOA formulation to three-dimensional space, addressing the scaling challenges inherent in aerial navigation.

  2. Vehicle Geometry: Transitioning from point-mass representations to a multi-point approximation of vehicle geometry, allowing for high-fidelity collision checking between 3D objects without excessive conservatism.

  3. Horizon Independence: Demonstrating RCOA’s unique property where the analytical formulation provides a guidance effect for obstacles residing outside the NMPC prediction horizon.

Performance Evaluation

The performance of RCOA is evaluated through two primary simulations involving quadrotor dynamics. The first assesses 3D navigation efficiency against notable OA benchmarks, demonstrating RCOA’s consistent computational advantage. p. 1

Where the paper says it

The performance of RCOA is evaluated through two primary simulations involving quadrotor dynamics. The first assesses 3D navigation efficiency against notable OA benchmarks, demonstrating RCOA’s consistent computational advantage.

The second highlights a high-performance maneuver navigating a narrow passage under aggressive roll angles. p. 1
Where the paper says it

The first assesses 3D navigation efficiency against notable OA benchmarks, demonstrating RCOA’s consistent computational advantage. The second highlights a high-performance maneuver navigating a narrow passage under aggressive roll angles.

Results show that RCOA facilitates successful navigation with a reduced prediction horizon, enabling controller frequencies over 30 Hz despite the nonconvex nature of the underlying dynamics. p. 1
Where the paper says it

The second highlights a high-performance maneuver navigating a narrow passage under aggressive roll angles. Results show that RCOA facilitates successful navigation with a reduced prediction horizon, enabling controller frequencies over 30 Hz despite the nonconvex nature of the underlying dynamics.

Comparative Analysis

The evaluation compares RNCOA against three standard OA formulations: the point to ellipsoid OA of (2) (EOA), the ellipsoid to ellipsoid overlap of (1) (EEOA), and the dual minimum distance between convex sets of (3) (DMDOA). RNCOA demonstrated unparalleled computational consistency, characterized by vastly narrower interquartile ranges compared to the other methods. p. 5

Where the paper says it

Each specific OCP configuration was executed 20 times to generate the statistical distributions illustrated in Figure 4. Across all topological corridors, RNCOA demonstrated unparalleled computational consistency, characterized by vastly narrower interquartile ranges.

While DMDOA and EEOA offer theoretical robustness in guaranteeing continuous non-intersection between two dynamic 3D bodies, this geometric fidelity incurs a prohibitive computational penalty that undermines real-time control viability. p. 5
Where the paper says it

Beyond mere consistency, RNCOA universally achieved the absolute lowest median computation times. While DMDOA and EEOA boast the theoretical robustness of guaranteeing continuous non-intersection between two dynamic 3D bodies, this geometric fidelity incurs a prohibitive computational penalty that undermines real-time control viability.

Narrow Passage Capability

In the narrow passage simulation, the prediction window was aggressively reduced to Tp = 0.25 s, utilizing a mere N = 9 temporal nodes. p. 7

Where the paper says it

0 (m) M1−4 50, 50, 3, 3 2) Narrow Passage Results: To actively evaluate the formulations under severe horizon constraints, the prediction window was heavily truncated to Tp = 0. 25 s, utilizing a mere N = 9 temporal nodes.

The RNCOA formulation successfully guided the vehicle even when the obstacle resided strictly outside the finite prediction horizon. p. 6
Where the paper says it

Highlighted: Configuration C navigating path A2 (green), Configuration D traversing A1 (blue), and Configuration B maneuvering through A4 (orange). Narrow Passage In this simulation, the prediction horizon is aggressively reduced to demonstrate two critical capabilities of the RNCOA framework: (1) its ability to successfully guide the vehicle even when the obstacle resides strictly outside the finite prediction horizon, and (2) its capacity to significantly reduce computational latency for high-frequency, real-time NMPC deployment.

This confirms that the continuous RCOA penalty acts as an anticipatory gradient, inherently regulating the optimal attitude well in advance of the hard spatial boundary. p. 7
Where the paper says it

Despite the obstacle remaining outside the solver’s look-ahead window in every iteration, each distinct sequence correctly actuates the required evasive roll angle. This visual evidence strongly confirms the theoretical assertion of Equation (10): the continuous RCOA penalty acts as an anticipatory gradient, inherently regulating the optimal attitude well in advance of the hard spatial boundary.

The resulting average solve time for the OA OCP during this maneuver was 11. p. 7
Where the paper says it

The computational statistics for the OA OCP during the simulated maneuver are provided in Table X. RNCOA maintained an average solve time of 11.

3 ms, firmly enabling sustained control frequencies exceeding 30 Hz. p. 7
Where the paper says it

RNCOA maintained an average solve time of 11. 3 ms, firmly enabling sustained control frequencies exceeding 30 Hz.

Conclusion and Future Directions

The paper concludes that RNCOA significantly outperforms standard baseline formulations in both raw computational speed and solver consistency. The formulation’s KKT derived continuous penalty gradient enables the NMPC to inherently anticipate obstacles and execute aggressive evasive maneuvers independently of the finite prediction horizon length. p. 9

Where the paper says it

Extensive numerical simulations demonstrated that RNCOA significantly outperforms standard baseline formulations, including ellipsoidal and dual set minimum distance constraints, in both raw computational speed and solver consistency. Most notably, the formulation’s KKT derived continuous penalty gradient enables the NMPC to inherently anticipate obstacles and execute aggressive evasive maneuvers (such as highroll passage navigation) independently of the finite prediction horizon length.

Future work will focus on generalizing the analytical properties of (10) to design customized penalty formulations that induce arbitrary optimal behavioral responses beyond standard Euclidean distance regulation. p. 9
Where the paper says it

Future work will focus on generalizing the analytical properties of (10). Specifically, we aim to design customized penalty formulations that induce arbitrary optimal behavioral responses beyond standard Euclidean distance regulation.

Additionally, the relaxed indicator logic of RCOA presents promising avenues for integration into broader Mixed Logical Dynamical (MLD) and hybrid control systems. p. 9
Where the paper says it

Specifically, we aim to design customized penalty formulations that induce arbitrary optimal behavioral responses beyond standard Euclidean distance regulation. Additionally, the underlying relaxed indicator logic of RCOA presents promising avenues for integration into broader Mixed Logical Dynamical (MLD) and hybrid control systems, potentially circumventing the need for computationally prohibitive MIP in state dependent logic transitions.

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

  1. Enhanced real-time maneuver execution by utilizing Relaxed Convex Obstacle Avoidance (RCOA) to enable a Nonlinear Model Predictive Controller (NMPC) to execute aggressive maneuvers through narrow passages with reduced prediction horizons, ensuring real-time feasibility at frequencies exceeding 30 Hz.

  2. Improved obstacle avoidance fidelity by transitioning from point-mass representations to a multi-point approximation of vehicle geometry, allowing for high-fidelity collision checking between 3D objects without excessive conservatism.

  3. Increased computational performance and consistency by demonstrating that RCOA provides computational performance on par or exceeding stateof-the-art methods when solving the Optimal Control Problem (OCP) for quadrotor dynamics.

  4. Enabled anticipatory guidance by leveraging the KKT derived continuous penalty gradient, which provides an intrinsic anticipatory guidance that regulates the entire boundary set safely above z o max, independent of the OCP prediction horizon.

  5. Improved maneuverability in constrained spaces by allowing the controller to execute aggressive evasive maneuvers even when obstacles reside outside the NMPC prediction horizon, breaking the dependency between prediction horizon and avoidance capability.

Abstract

The Relaxed Convex Obstacle Avoidance (RCOA) formulation is the first approach to enable a fully convex optimal control problem (OCP) for obstacle avoidance. Convergence analysis of RCOA yields an analytical framework that defines a unique characteristic: the ability to maintain obstacle avoidance (OA) efficacy even when obstacles reside beyond the controller's prediction horizon. In this paper, RCOA is extended to three-dimensional environments and apply it to Unmanned Aerial Vehicle (UAV) navigation. Furthermore, the formulation is enhanced to incorporate vehicle geometries, moving beyond point-mass representations to enable collision avoidance between 3D objects. Numerical simulations demonstrate that RCOA provides computational performance on par or exceeding state-of-the-art methods. Notably, RCOA is demonstrated to enable a Nonlinear Model Predictive Controller (NMPC) to execute aggressive maneuvers through narrow passages with reduced prediction horizons, ensuring real-time feasibility at frequencies exceeding 30 Hz.

Sources

Related papers