A team of researchers has designed a robotic system that allows an inexpensive, short-legged robot to navigate almost any obstacle or terrain. The robot can climb and descend stairs almost at its height or navigate rocky, slippery, rough, steep and varied terrain. He can also traverse holes, climb rocks, and operate in the dark.
The project The development of the system was carried out by researchers from the School of Computer Science at Carnegie Mellon University and the University of California, Berkeley.
Give little robots new skills
Deepak Pathak is an assistant professor at the Institute of Robotics.
“Enabling small robots to climb stairs and handle a variety of environments is crucial for developing robots that will be useful in people’s homes as well as for search and rescue operations,” Pathak said. “This system creates a robust and adaptable robot that could perform many daily tasks.”
The robot was tested on uneven stairs and slopes in public parks, which tested its ability to walk on stepping stones and on slippery surfaces. He was also tasked with climbing stairs which would be equivalent to a human jumping over an obstacle. The robot achieves an impressive ability to quickly adapt and master the terrain using its vision and a small on-board computer.
The robot was trained with 4,000 clones in a simulator. These clones practiced walking and climbing complex terrain, and the speed of the simulator allowed the robot to gain six years of experience in a single day.
Motor skills acquired during training were stored by the simulator in a neural network, which the researchers then copied to the real robot. This innovative approach meant that there was no manual engineering of robot movements.
Many robotic systems today rely on cameras that create a map of the surrounding environment, which is then used to plan robot movements before they are executed. However, this process can be slow and prone to errors due to inaccuracies or misperceptions during the mapping stage. These inaccuracies can have an impact on planning and travel.
While mapping and planning are useful for high-level control-focused systems, they’re not always best for the dynamic demands of low-level skills, such as walking or running.
Efficient and fast maneuver
The newly developed robotic system skips the mapping and planning phases and routes vision inputs directly to robot control. This basically means that the robot sees and moves accordingly. This revolutionary technique allows the robot to react very quickly and efficiently to its complex terrain.
The movements of the robot are driven by machine learning, which makes the robot inexpensive. The tested robot was at least 25 times cheaper than the alternatives on the market. According to the team, their algorithm could make low-cost robots much more accessible.
Ananye Agarwal holds an SCS Ph.D. machine learning student.
“This system uses vision and body feedback directly as input to send commands to the robot’s motors,” Agarwal said. “This technique allows the system to be very robust in the real world. If he slips down the stairs, he can get back up. He can go into unfamiliar surroundings and adapt.
The robotic system was strongly inspired by nature. For a robot less than a foot tall, it has learned to adopt the motions humans use to get over high obstacles in order to climb stairs or obstacles at its height. The system uses hip abduction to overcome obstacles that are difficult for even the most advanced legged robotic systems available.
The team also looked to four-legged animals for inspiration.
“Four-legged animals have a memory that allows their hind legs to follow the front legs. Our system works the same way,” Pathak said.
On-board memory allows the hind legs to remember what the camera has seen, helping it maneuver around obstacles.
Ashish Kumar holds a Ph.D. student at Berkeley.
“Since there’s no map, no planning, our system remembers the terrain and how it moved the front leg and translates it to the back leg, doing it quickly and perfectly,” says Kumar. .
The new research could play an important role in solving some of the key challenges surrounding legged robots. It might even help lead to their use in homes.
#lowcost #robot #overcomes #obstacles