Drawing inspiration from bats, researchers at Tel Aviv University have developed what they call a Robat, a fully autonomous terrestrial robot that, like a bat, emits sounds and analyzes the returning echoes to recognize, map and navigate obstacles outdoors. The development could have “great potential” for the use of sound for future robotic applications, the researchers said.
A study about the invention was published on Thursday in PLOS Computational Biology.
TAU graduate student Itamar Eliakim, together with a team of researchers in the fields of zoology, neuroscience and engineering, developed the robot, which, like a bat, emits sounds and analyzes the returning echoes to generate a map of space.
“One of the most challenging tasks, faced by many robots, is the problem of generating a map of an unknown environment, while simultaneously navigating through this environment for the first time,” the researchers, some of whom studied bat bio-sonar for more than 10 years, wrote in the study.
Bats solve the problem of mapping an unknown environment by perceiving their surroundings acoustically — emitting sound signals and analyzing the returning echoes, the researchers wrote.
Inspired by this ability, the researchers created the Robat — a robot that relies solely on a bat-like sound-based navigation system to orient itself through new environments and map them. To do this, the researchers used a biological approach, by creating ears — using two ultrasonic receivers — and a mouth, using an ultrasound speaker or emitter, which produced frequency-modulated chirps at a typical bat rate. This enabled the Robat to move through a large outdoor environment and map it in real time, they said.
To the best of their knowledge, the researchers wrote, “our Robat is the first fully autonomous, bat-like biologically plausible robot” that moves through a new environment “while mapping it solely based on echo information.”
This information helps delineate the borders of objects and the open space between them, said Eliakim, in a statement issued by the university.
The robot “achieved high mapping accuracy,” proving “the great potential” of using active wide-band sound emissions to map the environment and of using sound in future robotic applications, the researchers wrote. The researchers also created a machine learning algorithm that is fed by echoes from the environment to train the robot to better classify new objects.
The Robat was much slower than a real bat, stopping for some 30 seconds every half a meter to acquire the echoes, the researchers wrote. But this slowness was attributed by the researchers mainly to the mechanical limitations of the robot.
Any robot that needs to navigate its environment — and that is most robots — can benefit from the echo-based navigation algorithm developed by the study, Prof. Yossi Yovel of TAU’s Department of Zoology and a member of the research team said in an email to The Times of Israel.
This includes service robots, such as vacuum cleaner robots that need to navigate living rooms, agricultural robots that work in greenhouses, and rescue robots navigating under the ruins of a house that collapsed as a result of an earthquake.
The goal of the team was to “use nature in order to solve human problems,” Yovel, who is a bat expert, said. “Mimicking animals might not be the optimal way to solve these problems, for example a robot is not limited to two ears, but getting inspiration from animals can lead to new solutions.“
“Animals routinely solve problems that engineers find very difficult, like mapping a new environment while moving through it,” he added.
The project thus has two aims: on one hand it attempts to mimic animals to get a better understanding of the problems they face in reality. On the other hand, it seeks solutions to engineering problems that are currently very difficult to solve.
There have been several previous attempts globally to build a bat-like robots, Yovel said. But the Robat “is advanced in comparison to these attempts in several aspects.”
The Robat moves autonomously — previous robots were driven by a user — and it manages to delineate the borders of the objects that it detects. “When doing so it generates a map of open routes where it can move,” he said.
The Robat is also the first to use a machine learning algorithm fed by echoes from the environment to train it to classify new objects based on these examples, he said.
In future the researchers would like to mount their sensing unit on a flying robot. “This would not necessarily make it more accurate, but it would make it faster,” said Yovel.