A low-cost robot ready for any obstacle

Researchers at Carnegie Mellon University's School of Computer Science and the University of California, Berkeley have developed a new robotic system. This new robotic system enables a relatively small-legged robot at low-cost. This robotic system can do the following tasks:

• Climb and descend stairs nearly its height.
• Traverse rocky, slippery, uneven, steep, and varied terrain
• Walk across gaps; scale rocks and curbs
• Operate in the dark

Keywords: robotic system, small-legged robot, simulator,  motor skills, hand-engineering, Machine learning

Deepak Pathak, an assistant professor at the Robotics Institute said, "Empowering small robots to climb stairs and handle a variety of environments is crucial to developing robots that will be useful in people's homes as well as search-and-rescue operations. This system creates a robust and adaptable robot that could perform many everyday tasks". 

The team of researchers put the robot through various phases. They tested it on uneven stairs and hillsides at public parks. They challenged it to walk across stepping stones and over slippery surfaces. They analyzed this system by asking the robot to climb stairs at a height that is similar to leaping over a hurdle for a human. This robot is quickly adaptive and masters challenging terrain through its vision and a small onboard computer.

The team of researchers trained this robot with 4,000 clones of it in a simulator. They practiced walking and climbing on challenging terrain. The simulator's speed helped the robot in gaining six years of experience in a single day. The simulator also helped in storing the motor skills that the robot learned during its training. They stored it in a neural network and then copied it to the real robot. Compared to the traditional methods, this approach didn't require any hand-engineering of the robot's movements.

This new robotic system bypasses the mapping and planning phases and it directly routes the vision inputs for controlling the robot. The movement of the robot is determined based on what it sees. This technique helps the robot to react to the sudden oncoming terrain and move through it effectively. As there is no mapping and planning involved in this system and also it is trained using Machine learning, it is a low-cost system. The robot used by the team was at least 25 times cheaper than the available alternatives. 

Ashish Kumar a Ph.D. student at Berkeley said, "Since there's no map, no planning, our system remembers the terrain and how it moved the front leg and translates this to the rear leg, doing so quickly and flawlessly".

Ananya Agarwal, an SCS Ph.D. student in machine learning said, "This system uses vision and feedback from the body directly as input to output commands to the robot's motors. This technique allows the system to be very robust in the real world. If it slips on stairs, it can recover. It can go into unknown environments and adapt."

The team was also inspired by the movement of hind legs by four-legged animals. Researcher Pathak said, "Four-legged animals have a memory that enables their hind legs to track the front legs. Our system works in a similar fashion".

The development of this robotic system adds a new achievement in the field of robotics. It is a large step towards solving challenges faced by existing legged robots and bringing them to people's homes.