Humanoid robots with human-like bodies may soon help humans perform a wide range of tasks. For many robots, such tasks involve collecting objects of different shapes, weights, and sizes. While many modern humanoid robots are capable of lifting small and light objects, lifting bulky or heavy objects is often more difficult. In fact, if the object is too large or heavy, the robot can break or drop it. To address this problem, researchers at Johns Hopkins University and the National University of Singapore (NUS) have developed a method that allows robots to determine if they can lift a heavy box with unknown physical properties, TechXplore reports.
“We were particularly interested in how a humanoid robot could speculate about the possibility of lifting a box of unknown physical parameters,” Yuanfeng Han, one of the researchers who conducted the study, told TechXplore. “To accomplish such a complex task, the robot usually needs to first determine the physical parameters of the box and then create a safe and stable trajectory of movement of the whole body to lift the box.”
The process by which a robot generates motion paths that enable it to lift objects can be computationally intensive. In fact, humanoid robots usually have more freedom, and yet the movement that their body requires to lift an object must meet several restrictions. This means that if the box is too heavy or it’s center of mass is too far from the robot, it will most likely be unable to complete this movement.
“Think of us humans as we try to figure out if we can lift a heavy object like a dumbbell,” Han explained. – First, we interact with the dumbbell to get a certain sense of the object. Then, based on our previous experience, we kind of know if it is too heavy for us to lift or not. Likewise, our method begins by constructing a trajectory table that stores the various allowable lifting movements for the robot, corresponding to the physical range of the box using simulation. The robot then considers this table as information from its previous experience. ”
The technique, developed by Khan in collaboration with his colleague Ruixin Li and his leader Gregory S. Chirikjian (professor and head of the mechanical engineering department at NUS), allows the robot to gain an idea of the inertial parameters of the box after a brief interaction. The robot then revisits the trajectory table created by this method and checks if it includes a lifting motion that would allow it to lift the box with these calculated parameters.
If such a movement or trajectory exists, then lifting the box is considered possible and the robot can immediately complete the task. If not, then the robot considers the task to be beyond its capabilities.
“Our method can dramatically improve efficiency when performing practical pick and place tasks, especially if they are repetitive,” concludes Han. “In our future work, we plan to apply the approach to various objects or lifting tasks.”