Compact AI algorithm cuts machine automation setup times from hours to minutes.

Mitsubishi Electric Corporation announced that it has developed a proprietary deep-reinforcement algorithm for artificial-intelligence (AI) machine control. The compact algorithm requires just one-fiftieth the number of trials compared to conventional AI control methods by dramatically reducing the number of trials required for precise machine-learned AI control.

Conventional technology does not enable most equipment to achieve fully automated control due to the amount of data handling required. Instead, it requires human experts to provide teaching and programming input. Proprietary deep-reinforcement algorithms significantly reduce machine learning time using data from sensors and cameras. This dramatically reduces the number of trials and learning time required when compared to conventional deep-reinforcement learning methods.

The compactness of the innovative new algorithm means the new technology delivers AI systems at a much lower cost and with smaller server and network requirements compared to conventional systems. This ensures the AI can perform high-level inference even on embedded systems.

The algorithm is expected to enable smart equipment such as industrial robots and vehicles to use sensors and cameras to rapidly learn about their environments for finely tuned AI-based control in unique environments. Working in combination with Mitsubishi Electric’s Compact AI technology, the algorithm significantly reduces the calculation time compared to conventional methods, enabling deep-reinforcement learning to be deployed in a wide range of resource-limited equipment.

This enables machines with limited processing resources to use the solution to perform deep-reinforcement learning as it requires just one-hundredth the number of calculations compared to conventional methods. When comparing fully automated machine learning using the algorithm against existing machine learning supported by human experts, optimisation times can be reduced from hours to minutes.

Increasing industrial productivity.

The global market for AI is estimated to be worth US$ 35 billion, and expected to grow at 30 percent per annum as technology for fast, highly efficient automated machine learning that significantly reduces the amount of time and cost required to implement deep-learning AI control. Mitsubishi Electric’s deep-learning AI solution is expected to enable high-inference data processing by machines for increased industrial productivity.

 Mitsubishi Smart Automation at 999 Mitsubishi

See original press release