A team of roboticists from New York University has developed a robot that can perform tasks of finding and transporting objects in an unfamiliar environment. The article was published on: portal scientific publications arXive.
To accomplish such tasks, engineers used what is called a visual language model (VLM). It is based on the ability of a machine to recognize various objects based on linguistic cues, that is, to identify objects according to a specific description.
The researchers used a wheeled robot with an arm, called OK-Robot. During testing, the device was sent to the homes of 10 volunteers, where the droid was given a variety of tasks related to detecting and moving various things. For example, find a pink bottle and throw it in the trash. The challenge of the test was that OK-Robot had to follow instructions while navigating an unfamiliar environment.
Scientists asked the robot to perform 170 tasks. While the efficiency of the machine at the first stage of testing was 58%, it was later possible to increase it to 82%.
According to the authors of the development, the results demonstrate the feasibility of VLM-based robotic systems, as well as the possibility of using more complex robots.
engineers before was created The first robot roller for autonomous road repair.