When Engineering professor Jessy Grizzle first began the MABEL project in 2004, his intention was to use the bipedal robot as a platform to evaluate methods to achieve a dynamic sense of balance in legged machines. Since then, Grizzle said the robot has become more than just a mere platform.
With the ability to stand on two feet and move with external command, MABEL is one of the first robots capable of navigating through uneven terrain, Grizzle said.
In one demonstration, MABEL walked around its support pylon in a cyclical motion. To display its ability to respond to non-flat terrain, Koushil Sreenath and Hae Won Park — two Rackham students working with Grizzle on the project — placed a one-inch board in its path. MABEL stepped over the board and continued walking while Park and Sreenath continued placing additional boards in its path until the robot eventually stumbled and was forced to recover.
Grizzle said unlike other bipedal robots, MABEL is able to respond to uneven terrain without the help of a camera. He added that his team avoided implementing computer-based vision in MABEL to more closely emulate improvisational human behavior.
“The key feature is that MABEL has no camera and no visual sensors,” Park said. “It is like a human walking on rough ground blind.”
Sreenath said the mechanical aspects of MABEL have been modeled after human anatomy where the springs in the robot’s legs are analogous to tendons that absorb force. Members of the team said they worked closely with the School of Kinesiology to properly replicate the motion of the human legs.
“The tendons store energy so when the ground varies, the energy is stored in the tendons,” Sreenath said. “Equivalently, on MABEL we have springs … they compress to store energy so when the ground level changes, the energy does not propagate throughout the system.”
Grizzle said the team’s goal is to use MABEL to experiment on various algorithms that will help emulate human-like locomotion. He said they would like to look at what processes allow the human body to stay upright and move and translate them into an energy-efficient robot.
Despite the progress that has been made on MABEL, Grizzle said the project did not start out smoothly and when Sreenath and Park joined the team in 2007, MABEL was nowhere near operational. The team spent one year building the robot, which he said hindered the progress of Sreenath and Pak’s Ph.D. work.
“In the beginning, (Sreenath and Park) spent the first year of their Ph.D. work doing nothing but building a robot,” Grizzle said. “When it comes to a Ph.D. defense, that counts for zero … you’ve got to have results.”
Having recently overcome many of the setbacks that plagued MABEL’s construction, Sreenath said the team is now working on ways to improve the robot’s functionality.
With an electrical engineering background, Sreenath said he is concentrating on the electrical and control systems of MABEL and is looking to improve the accuracy of MABEL when it recalculates environmental factors like the coefficient of friction of the surface on which it is walking.
As a bipedal robot, MABEL is currently only able to walk. Park, who has a mechanical engineering background, said he hopes to eventually be able to increase MABEL’s walking speed and overcome the challenge of bringing both of the robot’s legs off the ground simultaneously. The result, he said, could be MABEL moving from a walk to a run.