Robots have transformed assembly lines in modern times, mastering repetitive tasks and moving with superhuman speed. Once the hard work of configuring and programming an industrial robot is complete, it will perform its job with perfection time and again. As long as you do not ask it to do a different job, that is. In the early days of the COVID-19 pandemic, this pain was felt as manufacturers repurposed some of their production lines, producing items like masks and ventilators instead of, for example, automobiles. Since the existing robots were not flexible in the tasks that they could perform, humans needed to man those assembly lines.
A major limiting factor for industrial robots is their grippers. These grippers are often passive to keep costs down, but that also limits their flexibility. Passive grippers cannot adapt to new situations. Researchers at the University of Washington have developed a method that allows for the production of inexpensive passive grippers and the associated control programs that allow robots to use them effectively. They claim that their technique makes it possible to quickly retool an assembly line such that it can accomplish a completely different objective.
To accomplish this goal, the team designed software that can take a 3D model of the object to be grabbed, and its position relative to the robot arm, as an input and turn that into instructions for a 3D printer to produce a custom gripper perfectly suited for the task. Another algorithm then searches for an optimal trajectory for the robot arm to approach and pick up the object — this is provided in the form of instructions that can be used to program the robot.
This method was evaluated by using a robotic arm to pick up 22 objects of a variety of shapes and sizes. During the experiment, each object was picked up 10 times using a custom, 3D-printed gripper and associated motion control plan. It was found that 16 of the 22 objects were correctly picked up in all 10 trials. Most of the remaining objects were picked up in at least some of the attempts, with the exception of two. In these cases, a bowl and a cup-like object were found to differ somewhat in form from the 3D models provided to the team’s software. You can hardly fault the method for the failures in these cases — garbage in, garbage out.
It was observed that the same gripping strategies were recommended for objects of similar shapes. The researchers took this as a positive indicator that it may be possible to create passive grippers that can pick up a range of similar objects. That could mean that new grippers do not necessarily need to be produced for each small variation in objects to be manipulated.
This technique may not work in all situations, however. Some objects are well suited to being manipulated by passive grippers, like those with protruding edges, for example. Objects with uniform, smooth edges, on the other hand, are very challenging to manipulate with any sort of passive gripper. Despite this limitation, this new method looks to be a very promising way to help assembly lines be quickly and inexpensively retooled.
