A sub kilogram transforming rover rolled out of an upside down lander, returned the first surface images of SLIM, and showed that distributed, low cost platforms can absorb the failures that flagship landers cannot.
When Japan's SLIM lander touched down on the moon on January 19, 2024, it did not land the way its engineers planned. The spacecraft came to rest nose-down on the lunar surface, its solar panels pointed away from the sun and its high-gain antenna aimed at the wrong part of the sky. Within hours, JAXA confirmed the obvious: SLIM had tipped over, and the lander would not survive a long lunar night.
It should have been a mission-ending failure. Instead, a palm-sized robot named LEV-2 separated from the lander moments after touchdown, rolled out on its own, and sent back the first surface images of SLIM on the moon. The data that little rover returned is the subject of a peer-reviewed paper in Science Robotics published on June 10, and it is the cleanest working proof yet that cheap, independent, autonomous platforms can rescue a science mission when the primary lander fails.
LEV-2, also known as SORA-Q, was a joint project between JAXA, Tomy, Sony Group, and Doshisha University. The robot weighs roughly 250 grams, fits inside a sphere about eight centimeters across, and transforms on command: it can split its shell into two wheel halves for rolling, or fold those halves into a butterfly-shaped stroke mode for short hops over uneven ground. The two modes are not a gimmick. They are an explicit answer to a problem lunar engineers have wrestled with for decades, which is that regolith is unpredictable, slopes are unknown, and any robot dropped onto a fresh surface has to cope with terrain its operators have never seen.
What makes the new Science Robotics account unusual is how directly LEV-2's design choices map to the failure mode it actually faced. The rover did not need to navigate a friendly, level landing zone. It needed to leave a tipped lander, avoid the shadowed body of the spacecraft, image an object lying on its side, and relay data back through LEV-1, a small hopper that also separated from SLIM at touchdown. The paper documents how the team built a robot with enough onboard autonomy to pick a path, switch modes when the surface got too rough, and complete its imaging run before the lander's thermal limits caught up with it.
That last point matters more than the cute-rover framing suggests. SLIM's power budget collapsed within hours of landing because its solar array was shadowed. The lander's electronics were designed to survive a single lunar day, roughly 14 Earth days, and only with sunlight. Once night fell, the lander went silent. Any science the mission was going to return had to come back before then, and the only platform capable of doing the work was LEV-2. The rover's job was not bonus science. It was the science return.
This is the design pattern worth watching, not the shape of the robot. Engineers planning the next generation of CLPS-era landers, ispace's lunar fleet, and any commercial payload headed to the south pole are now confronting a hard constraint: surface missions are getting cheaper, but the surface is not getting more forgiving. The bet behind LEV-2 is that you can absorb that risk by distributing it. A 250-gram robot that costs a fraction of a flagship instrument can be designed to fail forward. If the lander tips, the rover still rolls. If the rover stalls, the hopper still relays. If everything fails, the cost of the loss is bounded by the cost of the platform that failed.
The caveats are real. One mission is not a general autonomy proof. LEV-2's thermal and energy budget was tight enough that the team had to plan its imaging run down to the minute, and the robot did not survive the lunar night. The Science Robotics paper is honest about which parts of the run were autonomous and which required ground commands. LEV-2's success does not transfer to harder problems, like navigating permanently shadowed craters or surviving multi-month stays. What it does show is that a small, transforming, low-cost platform can do real engineering work under off-nominal conditions, and that the data it returns can be the most useful data the mission produces.
For lunar architects, the lesson is not "send more ball-shaped robots." It is that resilience can be a design parameter. The next decade of small lunar landers will fail in ways their teams cannot predict, and the question is no longer whether to plan for off-nominal landings but what to put on the lander that can keep working when the lander cannot. LEV-2, tumbling around an upside-down SLIM in the lunar dust, is the first published case study in how to answer that.