In a recent press release, the Mitsubishi Electric Corporation announced its development in artificial intelligence (AI) integrated automated guided vehicles (AGVs). The company has developed a cooperative AI technology that enhances work collaboration between humans and machines by using inverse reinforcement learning (IRL) to learn and imitate the actions of skilled workers.
IRL is a prime feature of Mitsubishi Electric’s Maisart (Mitsubishi Electric's AI creates the State-of-the-ART) AI technology, which is a machine learning technology that allows machines to mimic and learn human-like actions based on relatively small amounts of data. IRL is the field of learning an expert or skilled worker’s behavior by observing their rewards, objectives or values. The new cooperative AI technology will be further refined through the use of prototyping, persistent field testing by test deployment, and finally commissioning the technology into AGVs and robots alongside human workers at production and distribution sites. Mitsubishi Electric speculates that in the future, this technology will be used in autonomous driving vehicles and other applications.
Mitsubishi Shares the Key features of Cooperative AI for Human-Machine Work
This technology is expected to improve efficiency and safety in environments where both humans and machines work.
In sites where the environment involves humans and machines, AGVs will use the integrated AI technology to retrieve data from images in video recordings of these work areas to learn and imitate the actions and decision-making of skilled workers. In a mixed-work environment that features both humans and machines, this can help reduce dangerous and unwanted events such as high-risk actions, including collisions accidents or stalemates.
The results of in-house simulations conducted by Mitsubishi Electric showed increased operational efficiency by roughly 30 percent compared to conventional mixed-work environments that incorporate fewer intelligent machines.
As IRL requires relatively small amounts of data, the new Maisart technology reduces the amount of operational data required for learning.
Conventional machine learning is notorious for requiring an astoundingly large amount of data—in this kind of situation, video data, if the AI is to learn and imitate human actions. However, the development with using IRL reduces the amount of data required for the AI to learn and mimic human actions and decision-making. In recent testing by Mitsubishi Electric’s Maisart AI, the technology required only 10 percent or less of the video data that would be required by conventional machine learning algorithms.
Future of Cooperative AI for Human-Machine Work with Mitsubishi Electric
The project still requires work such as deployment and field testing. Moving forward, Mitsubishi Electric will continue to develop its new cooperative AI for possible applications in commercial facilities. The anticipated benefits include improved operational efficiency, enabling workers to maintain social distancing and allowing machines and humans to operate alongside each other safely in settings such as factory production lines and logistics warehouses, as well as applications for self-driving vehicles. Stay on the lookout of future updates and applications of IRL cooperative AI technology in AVGs.