Landing a rover on another planet has never been easy. Even at its best, the process can require careful choreography of multiple landing stages and involve parachutes, airbags, and retrorockets. Once you've landed, your wheeled spacecraft will face obstacles of its own, including steep terrain and sand traps.
Vytas SunSpiral and Adrian Agogino at NASA Ames Research Center in Moffett Field, Calif. and their colleagues suspect there might be a way to make solar system exploration much simpler and cheaper, by embedding science instruments inside a flexible, deformable robotic exoskeleton. This spherical structure might be able to land without assistance, absorbing most of the shock of impact itself and so saving mass needed for more complex landing gear. Once the spacecraft has reached an extraterrestrial surface, it could use this same structure to roll around without wheels, propelling itself by making slight tweaks to its shape.
The design, called the Super Ball Bot, relies on the concept of tensegrity, an approach to building structures that first emerged in the art world. Tensegrity structures consist of rigid components (such as hollow, cylindrical rods) connected by flexible materials (such as elastic cable). You may have already explored the basic properties of tensegrity structures with this classic baby toy. The term itself, coined by architect Buckminster Fuller, is a portmanteau of the words “tensional” and “integrity".
There is no single point of failure in a tensegrity robot, says SunSpiral, no axles or hinges that must be strengthened to withstand stress. Instead, the force of an impact—whether it's a landing or a fall off a cliff—is absorbed and diffused over multiple paths. "With a tensegrity structure, the entire structure shares the burden of reducing that stress, which is what you see in human bodies," he says. Since they are deformable, the robots could also interact more effectively with the environment, extricating themselves from surfaces, such as soft sand, that challenge traditional wheeled rovers.
But Mars, which has a thin atmosphere that would likely require a ball bot to carry a parachute, isn't the first target the team has in mind. Instead, they're considering Saturn's moon Titan, which boasts an atmosphere which can slow the spacecraft's descent enough so that the robot can be dropped, sans airbag or parachute, from 100 kilometers or more above the surface.
In a scenario studied by the team, the robot could be collapsed to a very compact configuration for launch. Once it reaches the moon, it would pop open and drop to the surface, flexing and absorbing the force of impact. By shortening and lengthening the cables that connect its rigid components, the ball bot could then roll about the surface. These same cables could be used to pull back parts of the robot, so that science instruments at the center could be exposed and used.
The team won a Phase I award from the NASA Innovative Advanced Concepts program to investigate the approach in 2012. Earlier this year, the team was awarded a Phase II grant to continue the research. You can hear SunSpiral and Agogino describe the basic approach and some of the work done so far in the video below.
One key challenge described in the video is the problem of control. The structures respond nonlinearly; a small change in the length of a connection can create a fairly big movement. They're also oscillatory – impacts and movements have a habit of reverberating through the structure. The problem is multifaceted and fast-moving enough that the team doesn't expect it to be practical for engineers to command the robot from Earth like they do the rigid Mars rovers, roughly one component at a time.
When I visited the lab the August before last, the team was just getting started on their research in an old aircraft control test facility on campus. Agogino told me that that they'd have to invent new ways to handle controls so the robot will move smoothly and efficiently. "This is so new," he said, that "if you can do anything more than a twitching dead thing you’re really ahead of the curve.”
To tackle the control problem, the team has used a physics simulator to model the interaction of the robot with simulated ground, on hills and undulating terrain, and with the addition of small obstacles. The simulated environment is used to identify the best approach in an evolutionary fashion, exploring many approaches to coordinating movement until an effective one is identified. (Others, incidentally, have used an evolutionary approach to create new robotic designs from scratch).
The team is also investigating more biologically-inspired control mechanisms that use oscillatory signals, similar to the central pattern generators (CPGs) that drive neural signals in animals. CPGs are already used for locomotion in multi-legged robots (see pdf).
In addition to the simulation effort, hardware prototypes (see photo at right) are being built to test controls and assess how well the structures protect fragile payloads (like uncooked eggs) from the force of a Titan landing.
As the video makes clear, there are still kinks to be worked out. The simulated ball bots perform much more smoothly than their real-world counterparts. But if all goes well, Titan might one day seen an armada of tumbleweed-like robotic explorers, hunting for evidence of life on the hazy moon. The approach, Agogino, says, “is unique. And it could be revolutionary.”