⚡ electrical engineering

B-spline shaping for low-thrust interplanetary rendezvous

This paper introduces a computationally efficient shape-based method for low-thrust interplanetary rendezvous that utilizes clamped B-splines to parameterize trajectories and algebraically recover control accelerations, thereby reducing the optimization problem to a finite-dimensional nonlinear program that consistently outperforms high-order hodographic shaping in terms of delta-v across various target scenarios.

Original authors: Julio C. Sanchez

Published 2026-07-28

📖 8 min read🧠 Deep dive

CC BY 4.0

Original authors: Julio C. Sanchez

Original paper licensed under CC BY 4.0 (http://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

Imagine you are trying to guide a tiny, super-efficient robot car across a vast, invisible ocean of space. This isn't a normal car with a gas pedal that you slam down for a quick burst of speed; it's a "low-thrust" vehicle, like a sailboat catching a gentle, constant breeze. It can't stop and start quickly. Instead, it has to push gently for months or even years to get from Earth to another planet. The big challenge for scientists is figuring out the perfect path for this slow, steady push. They need a route that uses the least amount of fuel (which is heavy and expensive to carry) while getting the job done in a reasonable time. If the path is wrong, the robot might run out of fuel before it arrives, or it might take so long that the mission fails. Finding this perfect "slow-motion" route is like solving a massive, three-dimensional puzzle where the pieces keep moving, and the solution needs to be found quickly so engineers can actually build the mission.

This paper introduces a clever new way to solve that puzzle using something called "B-spline shaping." Think of a B-spline as a flexible, digital ruler that engineers can bend into any smooth curve they want. Instead of trying to calculate every single tiny push the engine makes over years, the researchers use these flexible rulers to draw the entire path of the spacecraft first. Because of a special mathematical trick called "differential flatness," once they have the shape of the path, they can instantly figure out exactly how hard the engine needs to push at every moment to stay on that line. The paper tests this method against an older, more complex way of drawing paths (called "hodographic shaping") for trips to Mars, Mercury, a nearby asteroid, and a comet. The results suggest that this new "flexible ruler" method is not only faster to compute for standard configurations but also finds paths that use significantly less fuel, especially for the trickier destinations like Mercury and the comet. It's like finding a shortcut through a maze that everyone else was too busy calculating to notice.

The Problem: The Slow and Steady Space Race

Space travel usually gets a bad rap for being slow, but low-thrust engines are actually the champions of efficiency. Unlike chemical rockets that blast off with a massive, fiery explosion and then coast, electric thrusters (like those used on the Dawn and Hayabusa missions) provide a tiny, continuous push. It's the difference between a sprinter exploding out of the blocks and a marathon runner maintaining a perfect, steady pace. The problem is that because the push is so gentle, the spacecraft has to spiral around the Sun many times to reach its destination. This turns the mission design into a nightmare of math: you have to figure out exactly how to steer this slow-moving object for years without running out of fuel.

Traditionally, scientists have tried to solve this by breaking the journey into tiny steps and calculating the forces at each step, or by using complex "indirect" math that requires guessing the answer and refining it over and over. These methods are often slow, finicky, and can get stuck if the initial guess isn't perfect. The authors of this paper wanted a faster, more flexible way to design these trajectories, one that could quickly generate good solutions for different targets without needing a supercomputer to grind away for days.

The Solution: Drawing the Path with a Digital Ruler

The authors propose a method called B-spline shaping. To understand this, imagine you are drawing a smooth curve on a piece of paper. You don't draw the whole line at once; instead, you place a few "control points" (like pins on a board) and stretch a flexible strip (a spline) through them. By moving the pins, you change the shape of the curve.

In this paper, the researchers use a specific type of digital curve called a clamped B-spline. They use these curves to draw the path of the spacecraft in three dimensions: how far it is from the Sun, what angle it is at, and how high or low it is relative to the solar system's flat plane.

Here is the magic trick: The laws of physics for these low-thrust engines have a special property called differential flatness. This is a fancy way of saying that if you know the shape of the path and how fast it's changing, you can instantly calculate the exact force the engine needs to apply to stay on that path. You don't need to simulate the flight second-by-second. You just draw the path, and the math tells you the engine settings.

The researchers set up their "pins" (control points) so that the start and end of the curve perfectly match the position and speed of the Earth and the target planet. Then, they let a computer optimize the positions of the middle pins to find the path that uses the least fuel. Because the start and end are locked in by the math, the computer only has to worry about the middle, making the problem much smaller and faster to solve.

The Test Drive: Mars, Mercury, and Beyond

To see if this new method actually works, the team ran simulations for four different missions:

Earth to Mars: The classic red planet trip.
Earth to Mercury: A very hard trip because Mercury is deep in the Sun's gravity well.
Earth to Asteroid 1989 ML: A near-Earth asteroid.
Earth to Comet Tempel 1: A comet with a weird, tilted orbit.

They compared their B-spline method against the "high-order hodographic shaping" method, which is a well-known, high-quality technique that uses a different set of mathematical building blocks. They ran thousands of simulations, testing different departure dates and trip durations, creating what are known as "porkchop plots" (maps that show the fuel cost for every possible trip).

The Results: Faster and Leaner (With Caveats)

The results were impressive, but with some important nuances. Across the board, the B-spline method generally found paths that used less fuel than the traditional method, though the degree of improvement varied by destination.

Fuel Savings: The new method reduced the fuel needed (measured as $\Delta V$ , or change in velocity) significantly for the more difficult targets. For the trip to Mercury, the "median" fuel cost dropped by about 45.4%. For the asteroid 1989 ML, it dropped by 37.9%, and for Comet Tempel 1, by 31.6%. For Mars, the improvement was more modest; while the median fuel cost dropped by 21.8%, the absolute best (minimum) path found was only 1.4% better than the old method. This highlights that while the new method is consistently better on average, the "perfect" shortcut for Mars was already quite close to what the old method could find.
Speed: The speed of the method depends heavily on how complex the "ruler" is set up. The researchers found that their standard configuration (using 10 control points) was much faster, taking about 30–33% less time to compute each potential trip compared to the old method. A trip that took the old method 0.158 seconds took the new method only 0.109 seconds. However, if they increased the complexity to 40 control points to squeeze out extra fuel savings, the method became 11–13 times slower than the standard version. So, it's a trade-off: the simple version is a speed demon, while the complex version is a fuel-saver that takes much longer to run.
The "Sweet Spot": The researchers tested different numbers of "pins" (control points) on their digital ruler. They found that using 10 control points with a fifth-degree (quintic) curve was the perfect balance. It was fast, used very little fuel, and didn't require too much computing power. If they used 40 control points, they could squeeze out a tiny bit more fuel savings (especially for the best-case scenarios), but the computer had to work about 10 times longer to do it.

What This Means for the Future

The paper suggests that this B-spline approach is a powerful tool for the early stages of mission design. It allows engineers to quickly scan thousands of possible launch dates and find the best ones without getting bogged down in slow, complex calculations.

However, the authors are careful to note what this study doesn't do. They didn't include the weight of the fuel burning off (mass depletion) in their main calculations, and they didn't test the method on a real spacecraft. The solutions they found are "locally optimal," meaning they are the best in the specific mathematical family they tested, but they aren't guaranteed to be the absolute best possible path in the entire universe of math. Also, the method assumes the engine can push in any direction and with any strength, which isn't always true for real hardware.

Despite these limits, the study shows that by using a flexible, mathematical "ruler" to shape the path first, we can find better, faster, and cheaper ways to send our robotic explorers to the far corners of the solar system. It's a reminder that sometimes, the best way to solve a complex problem isn't to calculate every tiny detail, but to draw a better picture of the whole journey.

Technical Summary: B-spline Shaping for Low-Thrust Interplanetary Rendezvous

Problem Formulation
The paper addresses the preliminary design of fixed-time, minimum- $\Delta V$ low-thrust interplanetary rendezvous trajectories. The objective is to minimize the time integral of the control acceleration magnitude ( $\int_{t_0}^{t_f} \|u\| dt$ ) for a spacecraft traveling between two celestial bodies over a prescribed number of orbital revolutions ( $N$ ). The problem is formulated using heliocentric cylindrical coordinates ( $r, \theta, z$ ), where the boundary conditions (position and velocity) are derived from ephemerides at departure and arrival times. Unlike traditional optimal control approaches that require solving complex two-point boundary-value problems or large-scale nonlinear programs with dynamic constraints, this work seeks a parameterization that combines computational efficiency with sufficient solution quality for rapid preliminary design.

Methodology
The proposed method utilizes a shape-based approach leveraging the differential flatness of low-thrust cylindrical dynamics. The core innovation is the parameterization of the heliocentric cylindrical coordinates directly using clamped B-splines.

Differential Flatness: The cylindrical low-thrust dynamics are differentially flat with respect to the position coordinates $q = [r, \theta, z]^T$ . This property allows the control acceleration $u$ to be recovered algebraically from the trajectory and its first two derivatives ( $u \equiv u(q, \dot{q}, \ddot{q})$ ). Consequently, the equations of motion are satisfied by construction rather than through numerical propagation.
B-spline Parameterization: The trajectory $q(\tau)$ is represented as a linear combination of B-spline basis functions and control points. The independent variable is normalized time $\tau \in [0, 1]$ .
Clamped Boundary Conditions: By repeating the knots at the endpoints ( $\tau=0$ and $\tau=1$ ), the method analytically satisfies the position and velocity boundary conditions. The first two and last two control points are fixed to match the initial and final states, leaving only the interior control points as optimization variables.
Optimization Formulation: The continuous optimal control problem is reduced to a finite-dimensional nonlinear programming (NLP) problem. The decision variables are the free interior B-spline control points. The objective function ( $\Delta V$ ) is evaluated via numerical quadrature of the algebraically derived control magnitude. No dynamic collocation constraints are required.
Implementation: The NLP is solved using the IPOPT algorithm with a symbolic CasADi formulation. Initial guesses are generated by fitting the B-spline control points to a quintic polynomial profile in modified equinoctial elements.

Key Contributions
The paper makes two primary contributions:

A Novel Shape-Based Formulation: It combines clamped B-spline parameterization of cylindrical coordinates with the differential flatness of the dynamics. This allows for the algebraic recovery of control acceleration and the analytical satisfaction of boundary conditions, reducing the problem to a static NLP without numerical propagation of equations of motion. Unlike previous shape-based methods that construct problem-specific basis functions, this approach allows systematic adjustment of trajectory flexibility via standard B-spline hyperparameters (degree and number of control points).
Systematic Benchmarking: The method is rigorously tested against the high-order hodographic shaping method (a leading shape-based technique) across four distinct rendezvous scenarios: Earth to Mars, Mercury, the near-Earth asteroid 1989 ML, and Comet Tempel 1.

Results
Numerical campaigns were conducted on fine grids of departure dates and transfer times. The results demonstrate that the B-spline method consistently outperforms the high-order hodographic benchmark in terms of solution quality and, in specific configurations, computational efficiency.

Solution Quality: Across all tested targets and grid campaigns, all B-spline configurations reduced the mean, median, and minimum accumulated $\Delta V$ $Δ V$ relative to the hodographic benchmark.
- Mars: The quintic B-spline with 10 control points reduced the median $\Delta V$ by 21.8% and the minimum by 1.4%.
- Mercury: The quintic B-spline with 10 control points reduced the median $\Delta V$ by 45.4% and the minimum by 4.8%.
- 1989 ML: The quintic B-spline with 10 control points reduced the median $\Delta V$ by 37.9% and the minimum by 9.3%.
- Tempel 1: The quintic B-spline with 10 control points reduced the median $\Delta V$ by 31.6% and the minimum by 10.3%.
Computational Efficiency: The quintic B-spline with 10 control points was approximately 30–33% faster per optimization attempt than the hodographic method. While configurations with 40 control points yielded lower minimum $\Delta V$ values, they incurred significantly higher computational costs (7–13 times slower than the 10-point configuration).
Control Profiles: Higher-degree splines (quintic vs. cubic) and higher control point counts (40 vs. 10) produced smoother control transitions and more localized thrust arcs (impulsive-coast-like structures), respectively.

Significance and Claims
The paper claims that the proposed method provides a robust and efficient tool for the preliminary design of low-thrust interplanetary transfers. Specifically, it identifies the quintic B-spline with 10 control points as a favorable compromise between solution quality and computational efficiency, making it suitable for broad preliminary-design campaigns. The method's ability to handle diverse targets without manual construction of target-specific basis functions is highlighted as a significant advantage over previous shape-based techniques.

The authors modestly note that the results are limited to finite-dimensional shape spaces and an unconstrained acceleration problem (mass evolution and strict thrust limits were not included). The solutions are locally optimal within the chosen B-spline space and do not constitute proof of global optimality for the original continuous problem. However, the shaped trajectories are presented as effective rapid initial guesses for higher-fidelity direct or indirect optimal control methods.