Ackerman, Evan. “Legged Robots Do Surprisingly Well in Low Gravity.” IEEE Spectrum: Technology, Engineering, and Science News, IEEE, 28 June 2021, 18:00 GMT, spectrum.ieee.org/automaton/robotics/space-robots/legged-robots-surprisingly-well-low-gravity.
Article title: Legged Robots Do Surprisingly Well in Low Gravity
SpaceBok boks around on simulated asteroids, using its legs to keep oriented and under control
My summary of the article:
Researchers at the robotics lab in ETH Zurich are working on a very exciting project: making robots that jump on the surface on the Moon – or, not just the Moon, but other potential extraterrestrial places such as Mars or even asteroids. This idea originates from their realization that wheels, the conventional and undoubtably most 'convenient' means of transportation even used in Mars rovers, are not an efficient method of transportation in low gravity environments. This is mainly due to the fact that traction, or friction, a force that enables wheels to operate, is proportional to the object or robot's weight, which is directly dependent upon the gravitational field strength of that place. In asteroids, for example, with a gravitational pull or field strength of only 3% of that of the Earth, the friction between the surface and the wheel by the weight of the robot will not even be nearly enough to allow for the robot to travel around.
Thus, the researchers' solution to this problem was to replace wheels with legs. Inspired from the landing of cats from high altitudes, the robot that they have developed – named SpaceBok – is being designed to be able to operate by jumping and landing across the terrain of low gravity environments like the Moon or even asteroids. A problem that they have encountered while developing SpaceBok was its stability mid-air (or mid-flight since there is no 'air'). This is because only a mild disturbance in the initial jumping stage of the robot can cause severe changes to the motion of SpaceBok, especially in terms of its rotation. Of course, rotation is not ideal, since it may cause the robot to land not with its legs, and this may result in damage of the robot. Such mild disturbances cannot be prevented due to the fact that the surface textures of extraterrestrial objects (such as the Moon) are rough and unpredictable; thus, the researchers had to develop a way to ensure that SpaceBok can reorient itself into upright position mid-flight such that it can land on its four legs with no problem.
The researchers solved this problem by implementing deep reinforcement learning, a form of artificial intelligence. After developing their technique with deep reinforcement learning on a simulation, they transferred this technique into a real SpaceBok robot. According to the researchers, developing this technique was an extremely complicated process, since the robot had multiple limbs which made the system "highly non-linear", meaning that there were so many variables to control. At the end of the day, the researchers are hoping that SpaceBok could lead the research into extraterrestrial terrains in a much more efficient and timely fashion than robots that use wheels as their means of travel.
My response to the article:
There is a mythological creature that appears in Korean mythology called "Oktokki", or "Moon Rabbit". Legend says that this rabbit called Oktokki is pounding ricecake on the Moon, and this legend or myth comes originates from the appearance of the Moon's crater resembling a rabbit stirring or pounding a pot with a pair of sticks. Although this myth has nothing to do with the article, the fact that researchers are building a mechanical bouncing robot to send to the Moon reminded me of this legend.
This article perfectly addressed a curiosity I had with Mars rovers. Yes, they operate with wheels, and we use wheels for pretty much everything we build that travels, but how are we sure that technological choice to use wheels is the best choice on a different terrain with different gravitational field strengths? Surely, given that the functionality of wheels depend on the quantity of the force of friction between the wheels and the soil (which depends on the gravitational field strength or the weight of the object) and that Mars is much smaller planet than Earth, wheels may not be the best choice? Moreover, on terrains even smaller (or, precisely, lighter) than Mars, wheels would become an even less attractive choice.
It is evident that the researchers mentioned in this article had a similar doubt towards the effectiveness of wheels in lighter planets or bodies, and their solution is rather simple: attaching legs to robots and making them into basically hopping machines, or robot rabbits. Putting the robot itself aside, my biggest question for the team of researchers is how their tested the machine. They did mention that they used machine learning (artificial intelligence) in a simulated environment, but there must have been some real world adaptations before they are sure about the product; surely, a simulation is not enough to create a fully functioning algorithm. How were they able to mimic the environments of a lower-gravity celestial body without actually going there? Or did they just ignore that part and just relied on the simulation?
Another question that I have about this hopping robot is whether or not the algorithm is adapted to all gravitational field strengths weaker than that of Earth, or if it is just adapted to operation on a specific celestial body, say the Moon. This would be very important, and if SpaceBok is in fact adapted to all field strengths, I can barely imagine how many functions and operations SpaceBok will be able to operate in the near future. Who knows? Maybe one day we might have a space cat or a space rabbit delivering to us our space food.
Comments