The new cooperative AI technology will be further refined through test deployment in automated guided vehicles (AGVs) and robots at production and distribution sites where machines operate alongside humans.

Eventually, the technology is expected to be used in autonomous driving vehicles and other applications.

In mixed-work environments populated with humans and machines, Mitsubishi Electric’s collaborative AI technology enables AGVs to use images from video recordings of these work areas to learn and imitate the actions of humans. By learning actions such as yielding, the technology helps AGVs to avoid unwanted situations such as collisions or stalemates. In-house simulations conducted by Mitsubishi Electric raised operational efficiency by about 30 percent compared to operations in conventional mixed-work environments populated with less intelligent machines.

To enable AI to learn and imitate human actions, conventional machine learning requires huge amounts of operational data -in this case video data- which incurs time and cost burdens. Mitsubishi Electric’s Maisart AI, however, uses IRL to reduce the amount of data required to learn and imitate human actions. In simulations, the new technology required only 10 percent or less video data used normally.