New algorithms could improve safety of autonomous spacecraft

Throughout history, the sky has always seemed to humans like a vast empty space, a large empty dome punctuated during the day by the sun and at night by many small points of light (and periodically by the moon). As we venture into space, both physically, with spacecraft, and optically, with a range of telescopic technologies, we now know that there is a lot of stuff up there.

This discovery has profound implications for the aerospace industry. Imagine, for example, that a multibillion-dollar autonomous spacecraft, carefully designed and developed for years, is launched into space using precision calculations, only to lose one of its thrusters and crash into an asteroid.

Engineers have historically dealt with the possibility of equipment failure on spacecraft in two main ways: first, by having a “safe mode” in which the spacecraft can suffer minimal damage while scientists on the ground examine the data, make a diagnosis, and develop a solution; and second, by equipping autonomous vehicles with redundant systems. These allow a spacecraft, for example, to shut down a malfunctioning thruster and start using backup thrusters.

However, dangerous situations can arise in space without warning and without sufficient time to allow communications between space and Earth. And while redundant systems have proven very effective, they increase the cost and weight of autonomous spacecraft.

That’s why experiments are being conducted in the lab of Soon-Jo Chung, a professor of control and dynamical systems and a senior research scientist at JPL, which Caltech manages for NASA, to streamline the emergency functions of autonomous vehicles so they can diagnose and safely respond to encounters with other objects in real time. With new algorithms on board, spacecraft can test their own equipment and predict future actions that are most likely to keep them safe.

As one of the project’s supervisors, Fred Hadaegh, a research professor of aerospace at Caltech and a former technical lead at JPL, explains: “Having redundant systems is not always practical. It means the spacecraft has to be bigger, heavier, and more expensive than it otherwise would be. So the idea here is that when a spacecraft has a problem, it can figure out what’s wrong and fix or adapt to that specific defect.”

Chung’s lab houses, among other things, an advanced multi-spacecraft dynamics simulation facility.

“The simulator occupies a large room with a very flat floor,” says James Ragan, a doctoral student at the Graduate Aerospace Laboratories at the California Institute of Technology (GALCIT) and lead author of a new paper on the subject. “The model spacecraft uses air bearings to move along the floor with almost zero friction. At rest, it appears to float, and if you push it in one direction, it will keep moving until it hits something, which is the principle of space dynamics.”

Ragan programmed the robotic spacecraft simulator with what he and his co-authors call s-FEAST: Safe Fault Estimation via Active Sensing Tree Search. “Our s-FEAST algorithm rapidly ‘dreams’ of many possible futures that could result from the actions it takes now,” Ragan says.

“Because the system is noisy, these futures are uncertain. There are multiple possible outcomes, leading to a tree of branching possible futures. Each branch represents one possible way the future could play out, depending on things the spacecraft controls (the test actions it selects) and also things it doesn’t control, such as observations from faulty sensors.”

Chung adds: “What is innovative about our s-FEAST method is that we systematically solve the chicken-and-egg problem of estimating vehicle states, such as positions and velocities, and inferring failures or degradations, which are inherently coupled to each other.”

When the spacecraft detects unexpected data, it turns to the s-FEAST algorithm, which performs test actions “in the same way that you might carefully test your muscles when you feel unexpected pain and you want to figure out what hurts and how to avoid actions that might hurt you more,” Ragan says.

s-FEAST simultaneously develops a series of possible futures and selects from them the course of action that seems most likely to diagnose what happened and avoid the danger. In the case of this model, the danger is equivalent to a collision course with an asteroid.

“The key idea here is that s-FEAST doesn’t replace all of the spacecraft’s operations. It’s your emergency response,” Ragan explains. “The spacecraft gets an internal signal that something is wrong, so s-FEAST takes over all of the spacecraft’s computing power to quickly assess what’s going on and take corrective action. Once the hazard is identified and addressed, s-FEAST hands control back to the spacecraft’s regular computing environment.”

s-FEAST can also be used proactively. Suppose an autonomous spacecraft is about to undertake a particularly risky or mission-critical action; s-FEAST can run a test cycle to ensure that all systems are functioning properly before that action.

Chung and his co-authors believe the proposed method will establish a new way to make space exploration safer and more cost-effective. “Space systems make autonomous operations necessary because we can’t catch and repair Mars spacecraft and helicopters operating on a world far away from us,” Chung says. “Space is our ultimate ‘testing ground’ for any autonomy research we do for ground vehicle systems.”

Unsurprisingly, the s-FEAST algorithm that ran on the spacecraft simulator was adapted by the team to also run on a ground vehicle. Both experiments were successful, so s-FEAST technology holds great promise for autonomous vehicles on Earth as well as in space.

The research is published in the journal Scientific roboticsCo-authors are Ragan, Caltech postdoc Benjamin Rivière, Hadaegh and Chung.