Technologists from Rutgers University have deployed artificial intelligence to control a robotic arm so that it learns to find a more efficient way to pack boxes. This is a combination of hardware, 3D perception and robust motion.
While warehouses have begun to deploying robots in various forms, the act of shift through items and packing them carefully remains one of the main challenges in production robotics. The advantages with this form of automation are, however, considerable and will allow many companies to reduce their reliance upon manual labour.
To develop the new AI-driven robot, the researchers taught the robot how to move objects from a container into a shipping box and tightly arranging them, to enable secure transportation.
To help to teach the robot, the technologists used visual data plus a simple suction cup, which enabled to the robot to both lift and to push objects. The robot was also able to use sensor data to move objects toward a target area. The first wave of trials used cube-shaped objects, a second step would be look at objects of different shapes and sizes.
The following video shows the development of the robot:
According to the lead researcher, Kostas Bekris: “We can achieve low-cost, automated solutions that are easily deployable. The key is to make minimal but effective hardware choices and focus on robust algorithms and software.”
The development of the robot is described in a white paper, headed “Towards Robust Product Packing with a Minimalistic End-Effector.”
In similar news, engineers from Princeton University have constructed a robotic system that can assist in other picking and sorting tasks. This would also assist with warehouse sorting operations.
The “pick-and-place” robot if formed from an industrial robotic arm equipped with a custom gripper and suction cup. To facilitate the operation, the researchers built a “object-agnostic” grasping algorithm that allows the robot to assess a container of random objects and determine the best way to grip or suction any given item.