Several years ago, in order to improve working conditions and reduce the time required to complete tasks, the need for automation arose. Automation presupposes the existence of robots that actually “work” alongside humans. At first, the robots that were created performed some basic tasks and over the years, they were able to participate more actively in whatever was needed.
The problem with robots to date is that they can't predict where people are going. The result? They often stop moving altogether. The worst-case scenario is that they cause an accident by colliding with a human.
The problem is, while they could usually predict where a person was headed, due to poor time perception, the algorithms couldn't predict how long the person spent along their predicted path.
MIT researchers recognized the need to reduce this risk. To this end, they created an algorithm that better predicts the routes that people within a radius will take at any given time.
According to an MIT spokesperson, the algorithm is based on data that helps a robot understand and track human postures and movement. Existing algorithms predict trajectories based on distance, such as temporary stops where a person pauses before continuing on their path. The same goes for overlapping trajectories – cases where a person moves back and forth along a similar path.
Clearly, there is still a lot of work to be done before we can say that this algorithm works properly. The need to solve practical problems makes it at least necessary. Initially, factory and warehouse robots will be able to work closely with humans, without slowing down work or risking injury. MIT also said that robots can improve understanding of human behavior on other levels.
