We’re always impressed by the way David Zarrouk (a professor at Ben-Gurion University of the Negev by way of UC Berkeley’s Biomimetic Millisystems Lab) manages to extract a ton of functionality from the absolute minimum of hardware in his robots. In the past, we’ve seen clever designs like a steerable robot that only uses a single motor, and a multi-jointed robot arm that uses a traveling motor to actuate all of its degrees of freedom.
At the 2018 IEEE International Conference on Robotics and Automation (ICRA) in Brisbane, Zarrouk presented an update to STAR, the Sprawl-Tuned Autonomous Robot that we first wrote about in 2013. Called Rising STAR, or RSTAR, it takes the sprawling wheel-leg mobility and adds another degree of freedom that allows the body of the robot to move separately from the legs, changing its center of mass to help it climb over obstacles.
RSTAR is the latest in Zarrouk’s series of sprawling robots, designed to handle all kinds of terrain obstacles while minimizing cost of transport. “Sprawl” in this context refers to the robot’s legs, which are angled (adjustably) downwards and outwards from the body. RSTAR has an added degree of freedom in that its body is able to change its location relative to the legs, altering the robot’s center of mass.
It seems like a simple change, but it enables a bunch of new behaviors—not only can the robot climb over larger obstacles without flipping over, but it can also climb vertically up closely spaced walls and “crawl” through narrow gaps by adopting a legged walking gait.
While the adjustable center of mass helps keep the robot more stable, as the video shows flipping over can actually be useful, since it enables the robot to switch between faster and more efficient round wheels and more capable spoke wheels (whegs). RSTAR’s top speed is about 1 m/s on hard flat surfaces, although its turtle gait means that it can handle extremely soft or
granular surfaces (like thick mud or sand) without getting stuck.
At this point, none of what we’re seeing in the video is autonomous, although it sounds like they’re working on it—so for more details, we spoke with Zarrouk via email.
IEEE Spectrum: Why did you choose this particular modification for STAR to turn it into RSTAR?
David Zarrouk: We were looking to increase the capabilities of STAR in overcoming obstacles by adding a simple mechanism with one motor. We quickly converged to the concept of extending the distance between the wheels to the body. The question was whether we should use a mechanism that would or change the position of the center of mass (COM) in the fore-aft direction or not. The four bar extension mechanism (FBEM), which we eventually chose, moves the COM in the fore-aft direction making the robot even more dynamically reconfigurable. We eventually found that this interesting feature can also be used to increase stability or to intentionally pitch up or flip upside down when required to.
What are some situations or kinds of terrain that RSTAR can traverse that the previous version of STAR could not?
The original STAR is very capable in crawling underneath obstacles, over rough terrain and at very high speeds. But, similarly to all other robots, its climbing capabilities are limited by the size of its wheels. It can climb over obstacles that are up to 70 percent of its wheels’ diameter. By changing its height and width, RSTAR is more effective in running or rough terrain such as gravel, stones, or grass.
RSTAR can also crawl over granular or highly slippery terrain using the turtle gait without having to rotate its wheels. The height of the obstacles that RSTAR can climb is also substantially larger and mostly determined by the length of the rods of its FBEM. By moving its center of mass to the front, RSTAR can climb over steeper inclines without flipping over. RSTAR can also climb vertically in pipe-like environments and even crawl horizontally by pressing its wheels to the walls without touching the floor.
What are some potential applications for RSTAR or robots like it? Specifically, what potential applications would RSTAR be ideal for that other robots would have trouble with?
The STAR family of robots is very suitable to performing search and rescue operations, especially in unstructured environments such as collapsed buildings or flooded areas. In a genuine search and rescue operation, a robot often must overcome multiple successive obstacles of different types to reach its target. We built RSTAR having in mind that it should be simple, reliable and that it should be able to overcome multiple commonly available obstacles without any external mechanical intervention. RSTAR combines several climbing capabilities and shape changes which would allow it to climb over obstacles or sneak in between or under cracks. Besides that, RSTAR is relatively fast and has a relatively low energy consumption which together increase its working range and work time.
In what other ways would you like to improve RSTAR or make it more versatile?
There are two main upgrades that we currently have in mind for RSTAR. First, we are currently using machine learning algorithms for teaching RSTAR to perform some simple maneuvers to climb over specific obstacles. So eventually, RSTAR would have some partial autonomy. Second, we are considering increasing its mechanical capabilities by controlling the length of the FBEM rods.
Can you tell us more about the larger version of the robot at the end of the video?
As the size of obstacles a robot can climb is limited by its own size, we decided to build a larger version which would be able to climb over larger obstacles and stairs. The larger version can carry a 2 kg payload of sensors and supplies. It can also be used carry a smaller STAR or RSTAR robot. A larger STAR carrying a smaller STAR will increase the scouting capability of both, as the larger robot would carry the smaller robot over larger obstacles and closer to the destination, while the smaller one would be able sneak-in between smaller cracks and passages.