Solar Sail Interplanetary Transfer: Part II- Hybrid Layered Evolving Neural Network Based Global Optimization

Deep space travel has long been imagined as a journey fueled by massive tanks of chemical fuel, a method that limits how much cargo a spacecraft can carry and how far it can go. A more elegant alternative is the solar sail, a propulsion system that requires no fuel at all. Instead of burning propellant, a solar sail catches the gentle pressure of sunlight, much like a wind-powered boat catches the breeze. This continuous, renewable push allows a spacecraft to accelerate over time, potentially reaching high speeds and carrying larger payloads than traditional rockets. However, plotting a course for a solar sail is a formidable challenge. Because the sail must constantly adjust its angle to harness the Sun's light, finding the perfect path to another planet involves solving a complex mathematical puzzle with countless variables. Traditional methods often get stuck in local solutions, missing the truly best route because they rely on starting with a guess that is already very close to the answer.

A team of researchers from Malaysia, Japan, and Russia has developed a new way to solve this puzzle, one that combines the learning power of artificial intelligence with the trial-and-error strength of evolutionary algorithms. In a study focused on sending a solar sail from Earth to Mars, they created a hybrid system where a computer program acts as a learning agent. This agent uses a neural network, a type of computer model inspired by the human brain, to decide how to steer the sail at every moment of the journey. Unlike older methods that need a perfect starting guess, this system learns by evolving. It generates thousands of potential flight paths, testing them in a simulation to see which ones get closest to the target in the shortest time. The most successful paths are then refined further using a local search tool to polish the details. The researchers found that this approach could discover a flight path that was significantly faster than those found in previous studies, cutting the travel time to Mars by nearly nineteen percent compared to the highest estimates in existing literature.

The simulation demonstrated a journey where a solar sail, capable of generating a specific level of acceleration, left Earth and arrived at Mars in just under 365 days. This was a marked improvement over the 400 to 450 days typically predicted for such a mission using established methods. The system did not just find a faster route; it also determined the optimal launch date, pinpointing a window in mid-July 2020 as the ideal time to depart. Throughout the flight, the artificial intelligence continuously adjusted the sail's orientation, tilting it to catch the sunlight at the most efficient angles. The resulting trajectory was smooth and continuous, guiding the spacecraft to arrive near Mars with a relative speed low enough to be captured by the planet's gravity. The final distance between the spacecraft and Mars was within the planet's sphere of influence, and the relative speed remained well within safe limits for a rendezvous.

Beyond the flight path itself, the study offered a practical look at what kind of spacecraft could actually achieve this feat. The researchers analyzed the relationship between the size of the sail, the mass of the payload it could carry, and the performance of the mission. They determined that for a sail with a side length of roughly 100 meters, which is considered a realistic size for near-future technology, a payload of about 10 kilograms could be successfully delivered to Mars. This suggests that while the mission is demanding, it is within the realm of possibility for small, specialized probes. The study confirms that by using this hybrid neural network approach, engineers can navigate the complex, non-linear dynamics of solar sailing to find solutions that were previously out of reach, opening the door to more efficient and capable interplanetary missions.