OSARO has been granted a patent for its AI-powered robotic grasping technology, which enables robots to determine graspable object positions through physical interaction without relying on specialized sensors. The patent, titled “Computer-Automated Robot Grasp Depth Estimation” (USPTO #12,236,340), was issued on February 25, 2025, and describes methods by which pick-and-place robots collect interaction data to train machine-learning models capable of estimating object depth in various environments.
The technology is designed to enhance robotic handling in fulfillment warehouses, where inventory changes frequently, and grasping conditions can be complex. By integrating this patented depth-estimation software, part of the OSARO SightWorks™ platform, robots can adapt to new stock-keeping units (SKUs) without reprogramming, estimate depth across different objects, and grasp irregular or deformable items.
Gemma Ross, OSARO’s vice president of operations, stated that the innovation is particularly beneficial for kitting and bagging applications, where robots must handle a variety of items, including plush toys, reflective objects, and bagged apparel. The self-supervised learning process allows robots to improve their grasping accuracy by analyzing past attempts, reducing the need for human intervention in training machine-learning models.
Ross further explained that by repeatedly attempting to grasp an object and analyzing the resulting images and arm positions, robots can automatically refine their depth estimation capabilities. This approach reduces dependency on costly depth-sensing hardware, lowering overall system costs.
