⚡ electrical engineering

Generalised Mathematical Model for the Kinodynamic Planning of Multirotor UAVs and Aerial Robots using Topological Manifolds

This paper introduces a generalized mathematical model for multirotor UAV trajectory planning that treats the vehicle as a rigid planar point-set within topological manifolds, enabling configuration-aware kinodynamic planning that accounts for geometric constraints and actuator-level limitations often overlooked by conventional point-mass approaches.

Original authors: Yuvaraj George, Venkata Daseswara Rao Yendluri

Published 2026-10-09

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Yuvaraj George, Venkata Daseswara Rao Yendluri

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Flying machines that hover and zip through the air have become a familiar sight, from delivery drones buzzing over cities to autonomous robots inspecting bridges. To make these machines move safely and efficiently, engineers rely on mathematical models that predict how they will behave. For decades, the standard way to plan a drone's path has been to treat the entire vehicle as a single, weightless dot floating in space. This "point-mass" approach ignores the fact that a real drone has a physical shape, with motors and arms sticking out in specific directions. While this simplification made early calculations possible, it assumes that if the center of the drone can fit through a gap, the whole machine can follow. In the real world, however, a drone's size and the fixed distance between its spinning propellers matter just as much as its center point. If a drone is too wide, it will crash into a wall even if its center could theoretically pass through.

Researchers at the Birla Institute of Technology and Science in Hyderabad have developed a new mathematical framework that fixes this oversight. Instead of treating a drone as a single dot, they model it as a rigid, flat shape made of several fixed points, representing the actual positions of the motors. This approach allows them to track not just where the center of the drone goes, but exactly how every single motor moves as the vehicle tilts, turns, and travels. By doing this, they can see if a planned path is truly possible for a specific drone design, rather than just assuming it works for any drone that fits the same center-point path. Their work, published as a research article, suggests that ignoring the physical dimensions of a drone can lead to plans that look perfect on a computer screen but are impossible to fly in reality.

The core of this new method involves mapping the drone's movement onto a "topological manifold," which is essentially a mathematical surface that preserves the fixed distances between the drone's parts. Imagine a drone as a flat, rigid plate with four motors attached at the corners. No matter how the drone spins or flies, the distance between those corners never changes. The researchers built a system that respects this rule at every step of the journey. They created equations that take a smooth path for the center of the drone and automatically calculate the exact path each of the four motors must follow to keep the plate rigid. This means that if the drone tilts to the side, the model instantly knows that the motor on the left must move slightly differently than the one on the right, even though the center of the drone is moving in a straight line.

In their study, the team tested this idea by simulating a drone flying through a specific mission profile. The drone started on the ground, climbed to an altitude of 50 meters, traveled 200 meters forward, and then descended. They compared their new method against the traditional point-mass approach. The results showed that while the center of the drone followed the exact same path in both models, the individual motors behaved very differently. In the new model, the motors traced unique, complex curves that accounted for the drone's rotation and physical size. The researchers found that the speed, acceleration, and sudden changes in force (known as jerk) at the motor level were significantly different from what the point-mass model predicted. This difference is crucial because the motors are the parts that actually push the drone; if they are asked to move in a way that exceeds their physical limits, the drone will fail.

To prove the importance of this distinction, the researchers simulated a scenario where two drones with different sizes were given the same flight path. Both drones were designed to fly through a narrow cylindrical pipe. The traditional point-mass model, which only looks at the center, predicted that both drones could make the trip because their centers were small enough to fit. However, the new model revealed that the larger drone would crash into the walls of the pipe. Even though its center could pass through, its wider body and the fixed distance between its motors meant it physically could not fit. This simulation highlighted a critical flaw in the old method: it cannot account for the physical bulk of the vehicle. The new framework, by keeping track of the actual shape, correctly identified that the larger drone's path was impossible, preventing a potential crash before it happened.

The study also explored how the size of the drone affects the smoothness of its flight. The researchers ran simulations with drones of varying sizes, from a small three-motor design to a large eight-motor version, all following the same center-point path. They measured the "curvature," or how sharply the path bends, for the motors on each drone. They found that as the drone got larger, the paths traced by the motors became more curved and demanding, even though the center of the drone moved in the same smooth way. For the smallest drone, the motors followed a relatively gentle curve. For the largest drone, the motors had to trace much sharper turns. This means that a flight plan that is easy for a small drone might be too difficult for a larger one, simply because the outer parts of the larger machine have to swing through a wider arc. The new model captures this reality, allowing engineers to check if a specific drone design can actually handle a planned route.

By treating the drone as a collection of fixed points rather than a single dot, the researchers have created a tool that bridges the gap between high-level planning and low-level control. This approach ensures that the commands sent to the motors are physically realizable. It removes the need for complex, error-prone conversions between different ways of describing rotation, which often cause mathematical errors known as singularities. Instead, the system works entirely with standard three-dimensional coordinates, making it more robust and easier to implement on the computers that fly these machines. The researchers demonstrated that this method works for any number of motors, whether the drone has three, four, six, or eight, making it a versatile solution for the growing variety of aerial robots being developed today.

The implications of this work extend to the safety and reliability of autonomous flight. As drones are deployed in complex environments like warehouses, forests, or urban canyons, the margin for error shrinks. A planning system that ignores the physical size of the vehicle risks sending it into a collision. The new framework provides a way to verify that a path is not just theoretically possible for a point, but physically possible for the actual machine. The researchers suggest that this method can be integrated into existing flight control systems as an additional layer of safety, checking constraints in real-time without needing to change the fundamental way the drone is controlled. While the current work relies on computer simulations, the mathematical foundation is solid, and the team plans to test these ideas on real hardware in the future.

Ultimately, this research shifts the perspective on how we think about flying robots. It moves away from the convenient but inaccurate idea of the drone as a ghostly point and embraces the reality of the drone as a solid object with a specific shape. By doing so, it ensures that the paths we plan are paths the machines can actually fly. The findings suggest that for autonomous drones to operate safely in the real world, their planning software must understand their own geometry. The ability to predict exactly how every part of the machine will move, based on its physical dimensions, is a significant step toward making autonomous flight more reliable and capable.

Technical Summary: Generalised Mathematical Model for the Kinodynamic Planning of Multirotor UAVs using Topological Manifolds

Problem Statement
Conventional trajectory planning for Multirotor Unmanned Aerial Vehicles (MUAVs) typically relies on a "point-mass" abstraction, where the vehicle is modeled as a single point at its Center of Gravity (CG). While computationally efficient, this approach fails to explicitly account for the vehicle's physical dimensions, rigid-body geometry, and fixed inter-rotor distances (holonomic constraints). Consequently, point-mass models may generate trajectories that are kinematically feasible for the CG but physically unrealizable for the actuators (rotors) due to geometric constraints. This "paradox of point-mass kinodynamic planning" can lead to feasibility gaps where a path deemed admissible for a smaller vehicle configuration becomes collision-prone or kinematically invalid for a larger configuration sharing the same CG trajectory. Furthermore, existing generalized models often sacrifice topological consistency or rely on specific configurations, limiting their applicability across diverse MUAV geometries.

Methodology
The authors propose a generalized mathematical framework that models the MUAV not as a point mass, but as a rigid planar point-set (referred to as a "rotor polygon") embedded in $\mathbb{R}^3$ . This approach utilizes topological manifolds to represent the vehicle's configuration space.

Geometric Representation: The vehicle is defined by $n$ vertices (motors) forming a symmetric, closed, convex polygon. The model enforces holonomic constraints where the distance between any vertex and the CG remains constant ( $r$ ), preserving the rigid-body structure.
Unified Kinematic Formulation: The framework employs homogeneous transformation matrices to map the CG's position and orientation (Euler angles) to the global Cartesian coordinates of all $n$ vertices. This creates a unified Euclidean representation where the motion of individual rotors is expressed within the same Cartesian framework as the CG.
Forward and Inverse Kinematics:
- Forward Kinematics: Given the CG trajectory and orientation, the model computes the exact 3D positions, velocities, and accelerations of all rotor vertices.
- Inverse Kinematics: The authors formulate a cubic parametric kinodynamic model (third-order polynomials) to generate time-continuous trajectories. This approach solves for trajectory coefficients using boundary conditions (position, velocity, acceleration) at specific time intervals, ensuring smoothness and continuity without the computational overhead of higher-order polynomials (e.g., quintic or minimum-snap).
Constraint Handling: The model implicitly incorporates dynamic bounds (velocity, acceleration, jerk) as boundary conditions rather than explicit dynamic optimization, allowing for real-time implementation while ensuring trajectories remain within actuator limits.

Key Contributions

Generalized Rigid-Body Model: A mathematical model representing MUAVs as rigid planar point-sets in $\mathbb{R}^3$ , explicitly enforcing fixed inter-point distance constraints while maintaining standard rigid-body degrees of freedom.
Actuator-Level Propagation: A unified kinematic formulation that enables the direct propagation of CG trajectories to actuator-level (vertex) trajectories. This allows for the evaluation of curvature, acceleration, and jerk at the individual rotor level, which is impossible with standard point-mass planning.
Configuration-Aware Planning: A framework that extends conventional planning by incorporating physical dimensions into feasibility and admissibility evaluations without altering the underlying rigid-body motion representation. It does not replace existing path planning algorithms but augments them with kinematic feasibility metrics for individual rotors.
Singularity Avoidance: By operating primarily in Cartesian coordinates, the inverse kinematic solution avoids singularities commonly associated with angle-based parameterizations in both control and kinematic computations, ensuring smooth and continuous motion generation.

Results and Simulation
The authors conducted simulation studies comparing the proposed topological manifold approach against conventional point-mass planning across various multirotor configurations (tricopter, pentacopter, hexacopter, octacopter) with varying rotor arm lengths ( $r$ ).

Identical CG, Divergent Actuator Profiles: Simulations demonstrated that while different configurations can share an identical CG trajectory, their actuator-level kinematic profiles (curvature, acceleration, and jerk) differ significantly.
Curvature Amplification: Results showed that as the rotor arm length ( $r$ ) increases, the maximum vertex curvature increases monotonically. For instance, in a fixed CG trajectory, the maximum vertex curvature rose from 0.53651 (tricopter, $r=0.30$ m) to 0.63956 (octacopter, $r=0.88$ m).
Theorem Validation: The study validates Theorem 1, which posits that a collision-free path admissible for a smaller rotor polygon ( $r_A$ ) is only guaranteed to be admissible for a larger polygon ( $r_B$ ) if $r_B \leq r_A$ . This highlights that point-mass planning can falsely predict feasibility for larger vehicles in geometrically constrained environments (e.g., narrow pipes).
Feasibility Metrics: The proposed model successfully generated unique control profiles for each of the $n$ motors, demonstrating that the system can resolve the over-actuation or under-actuation issues by treating each vertex as a distinct state variable within the manifold.

Significance and Claims
The paper claims that this framework addresses a critical gap in MUAV autonomy: the disconnect between high-level path planning (often point-mass based) and low-level actuator realizability. By integrating physical dimensions directly into the planning formulation via topological manifolds, the model provides a configuration-aware extension to conventional planning.

The authors assert that this approach:

Enhances the completeness of kinodynamic path planning by explicitly addressing the relationship between physical dimensions and path feasibility.
Provides a systematic mechanism for handling actuator redundancy and geometric constraints without requiring full dynamic optimization.
Offers a robust foundation for both single-vehicle and coordinated multi-vehicle operations where geometric constraints cannot be ignored.
Avoids the need for complex post-processing to map CG trajectories to actuator commands, as the mapping is intrinsic to the mathematical formulation.

The work concludes that while point-mass models remain useful for initial decision-making, the proposed manifold-based approach is necessary for ensuring that planned trajectories are physically realizable and kinematically feasible for the specific geometry of the aerial robot.

Technical Summary: Generalised Mathematical Model for the Kinodynamic Planning of Multirotor UAVs using Topological Manifolds

More like this