Digit Geek
Digit Geek > Recent Articles > Technology > Humanoid locomotion in robots

Humanoid locomotion in robots

The promises of robotics is to reduce the need for humans to do repeated, menial or dangerous jobs. Applications range from agriculture to manufacturing to transport. For many of these applications, it is not necessary for the robots to resemble the humans. The robots can grow, shrink, crawl, fly, roll, sail or glide in their efforts to navigate the environment and perform the task at hand. Shapeshifting robots can even shift from one form to another. The arms used to manipulate objects do not need to look like human limbs. Research and current engineering trends suggest that in the future, robots will not resemble humans at all. This is because humans are simply too inefficient. Dr Simon Watson, a lecturer in Robotic Systems at the University of Manchester says, “Humanoid robots are a vanity project: an attempt to create artificial life in our own image – essentially trying to play God. The problem is, we’re not very good at it. Ask someone on the street to name a robot and you might hear “Terminator”, “the Cybermen” or “that gold one from Star Wars”.

The “gold one from Star Wars” is a protocol droid, designed to interact with organic life forms. Image: StarWars.com.

There are reasons we do need androids, or robots that resemble humans in appearance. One reason is for providing assistance to those with impaired mobility, or disabilities. The technologies developed for these kinds of robots can also have applications in prosthetics. Robots that care for the elderly, young children, or for use in the service industry stand to benefit by appearing human, even more so if they are cute. While the human form factor might not be very efficient when it comes to doing one kind of task, it is pretty versatile.

In the immediate aftermath of the Fukushima Daiichi nuclear disaster, humans could not go back into the plant because of how dangerous the area was. However, if a robot entered the plant and the cooling system back on, the extent of the damage caused by the reactor meltdown could have been significantly reduced. The disaster urged DARPA to spur innovation towards a robot that could have mitigated the disaster. The DARPA robotics challenge had a series of tasks for robots to perform, including driving a vehicle, climbing a ladder, opening a door, connect a hose pipe and breaking through a concrete panel. While some robots in the challenge used tank tracks or had four legs, most were bipedal.

Not a walk in the park

We really are not good at building bipedal robots. Robots are not even smart enough to stand up. The most advanced humanoid robots tend to fall over when performing simple tasks such as climbing stairs, opening a door, or even walking. This is difficult to accomplish when the robots are given four legs. A significant portion of the research into humanoid robotics is dedicated to just keeping them upright. In fact, researchers at Georgia Tech are looking into training robots to fall gracefully, in a manner that causes as little damage as possible to the robot, as well as others. This can help reduce the costs of repair, and is an important skill when it comes to robots working with the injured, children, elderly or pets.

The biomimetics robotic laboratory at MIT is developing a robot called HERMES to respond to disaster situations. The idea is to keep a human in the loop, controlling the robot remotely, and provide a full body teleoperation platform. While performing the actions, such as using an axe to break through a wall, the robot can easily fall after performing the action once. A metal exoskeleton attached to the waist translates information from sensors on the robot specifically designed to measure the changes in the weight distribution of the robot. The human operator can feel the shift in weight, and adjust their stance accordingly, letting the robot perform momentum driven tasks. Through the use of the bilateral feedback, the robot moves just like the human in real time.

An approach by researchers at the University of Purdue was to use a Kinect camera from Microsoft to record a human performing actions, and then map the actions to a Hubo II, an android manufactured by Rainbow Robotics. Through this approach, the robot was taught the graceful motions of Tai Chi, as well as to dance to Gangnam Style. While the arm movements were accurate renderings, the robot was not able to move its legs in the way that Psy does.

Most androids lean forward at the waist and bend their knees to keep their balance while walking. The gait is not human like, as the robots tend to place the entire foot on the ground at once, before moving the other leg. Researchers from Georgia Tech have given a robot a more natural gait by allowing it to place its heels first on the ground, then roll forward, and push off with the tip of its toes. This gait resembles the human manner of walking more, allowing the team to significantly reduce the amount of energy it takes for the robot to walk. The robot was even given a pair of shoes meant for humans, and the next step for the team is to make the robot climb stairs, and run across fields.

The Atlas platform by Boston Dynamics is one of the most advanced humanoid robots in existence. The robot balances by taking the input from multiple sensors in its body and legs. Atlas was used by no less than six teams appearing in the DARPA robotics challenge finals. Even though Atlas has bent knees and a flat footed gait, it can perform carry out tasks. The robot can navigate uneven terrain, stand on one leg, jump between platforms, perform backflips and yes, even open doors.

Getting a grip

Robotic hands have traditionally been problematic for a number of reasons. They are not nearly as dextrous and flexible as human hands. The human hand itself is anatomically one of the most complicated areas in the human body. While robotic hands can replicate this complexity by using a number of sensors and motors, these kinds of hands would be complicated to program, and so expensive that they would be impractical. Researchers from Yale and Harvard approached this problem by taking inspiration from another creature – the humble cockroach. The flexible and springy legs of the cockroach allows them to navigate tricky surfaces, and using the same approach for robotic graspers also increased their efficiency at holding objects of various shapes and materials. The advantage of this approach is there are fewer motors and sensors, reducing the weight of the robotic hand for use in prosthetics.

A robotic hand with grippers inspired by cockroach legs. Image: William Sacco, Yale University

One promising area of research that can have solutions is the field of soft robotics. The cold, hard hands of robots are not very great at say, delicately handling pets. Researchers from Harvard have developed a soft robotic glove that can help patients with loss of motor abilities in their hands. The robot is just like a glove, and fits easily over a hand, improving the gripping capability of the patient. The glove is made up of soft multi segment actuators, that supports the entire range of motions possible with the human hand.

Soft Robotic Glove from Wyss Institute on Vimeo.

Engineers at Riken, a research institute in Japan, in collaboration with Sumitomo Riko Company have developed the Robear, a medical robot with the appearance of a teddy bear. The feedback from the actuators in realtime, while performing the tasks, allows the robot to handle patients gently, even when performing power intensive tasks, such as lifting them. The legs have extensions that come out when carrying a patient, for stability, but retract when the robot is navigating tight spaces.

A challenge that robotics in prosthetics is to allow the human brain to manipulate the appendages. Researchers from Duke University have successfully demonstrated monkeys manipulating robotic arms using signals captured from their brains. Researchers from MIT and Boston University have developed a system that allows a human to correct mistakes made by a robot through an ECG interface. An AI process the brain activity recorded by the ECG monitor and passes on the instructions to the robot. Such research in the future could allow for the development of neuroprosthetics, and independent robots that can be controlled with the brain.

Researchers from Georgia Tech have experimented with a smart arm for a drummer. Attached to the shoulder, it allows the drummer to actually have three arms. The robot responds to human gestures, as well as the music that it hears. The arm keeps speed with the tempo of the drummer, and dynamically moves to parts of the drum kit that the drummer is not playing. Robotic arms can do much more than just restore lost capabilities, and can augment existing capabilities.

Aditya Madanapalle

Aditya Madanapalle

An avid reader of the magazine, who ended up working at Digit after studying journalism, game design and ancient runes. When not egging on arguments in the Digit forum, can be found playing with LEGO sets meant for 9 to 14-year-olds.