The development of physical artificial intelligence could open up opportunities for Armenia not only in training specialists and conducting scientific research, but also in building its own robotics industry, attracting international technology companies, and developing solutions for industry, emergency response, and other fields. However, this will require addressing several challenges at once: a shortage of qualified personnel, limited computing resources, an insufficient number of research laboratories, and the need to bring together research groups that currently work in isolation.

The seven-year program for the development of physical artificial intelligence approved by the government is intended to address these challenges.

From Separate Research Efforts to a Unified Scientific Ecosystem

Speaking about the program's impact on the work of the Laboratory of Intelligent Systems and Advanced Control Systems at the National Polytechnic University, laboratory head and American University of Armenia lecturer Astghik Hakobyan noted that physical artificial intelligence is a multilayered, interdisciplinary field whose development requires more than expertise in robotics and control systems alone. It requires bringing together specialists from various fields, including computer vision, data processing, environmental perception, and other areas essential to creating autonomous intelligent systems.

Armenia already has research groups working in some of these areas. However, their activities have not yet been integrated into a unified research framework focused on physical artificial intelligence.

“We need to find a way to bring these groups together around physical artificial intelligence. That is exactly what we intend to do through this program,” Hakobyan explained.

According to her, the Laboratory of Intelligent and Advanced Control Systems at the National Polytechnic University already has modern equipment. However, tackling more complex research tasks requires additional hardware resources, computing equipment, and, just as importantly, human expertise. Modernizing the laboratory therefore involves more than simply purchasing new equipment. It means expanding research capabilities, attracting specialists from related fields, and creating conditions for research teams to work together.

This is particularly important for physical AI: developing a robot capable of navigating its surroundings independently and carrying out assigned tasks requires simultaneous work on software models, control systems, sensors, information processing, and safety.

YSU: Existing Computing Capacity Is No Longer Sufficient

Hrant Khachatrian, head of the Machine Learning Laboratory at YSU's Center for Mathematical and Applied Research and the YerevaNN laboratory, also addressed the issue. According to him, YSU already has research groups working on artificial intelligence and receiving funding through various grant programs run by the Higher Education and Science Committee.

These groups work in different areas of AI, and over the past year or two, more than half have gradually begun shifting their research focus toward the physical world. As an example, Khachatrian cited the development of artificial intelligence models designed to control drones. Such systems must not only process information but also make decisions that directly affect an aircraft's movement and actions.

For now, however, these developments remain at the computer simulation stage. Researchers cannot yet deploy the models they have developed on real drones and use them in practical settings with sufficient confidence.

“For now, we are doing everything through computer simulations, until we are confident that the systems are safe, reliable, and accurate enough to be tested on real drones,” he explained.

This is a crucial stage in developing physical AI systems. A model that works successfully in a virtual environment will not necessarily perform just as effectively in the real world. Interaction with the physical environment introduces additional factors: changing conditions, unforeseen obstacles, perception errors, and the need to respond quickly to events.

The transition from simulations to real-world testing therefore requires not only more advanced algorithms, but also substantial computing resources, equipment, and specialists.

Khachatrian pointed out that YSU already operates a large supercomputer, but it is heavily used, and the available computing capacity is no longer sufficient for researchers to carry out their research tasks in full.

A shortage of personnel remains another challenge.

The research groups associated with his work employ a total of around 20 scientists. Most are young specialists with relatively limited research experience so far. This number of researchers is insufficient to achieve significant results at the international level in such a competitive field.

“We need additional resources to expand our teams,” Khachatrian emphasized.

One solution he identified was postdoctoral programs, which would make it possible to attract scientists who have completed doctoral research at universities abroad and already have experience conducting independent research. Their participation would give Armenian research groups access to additional knowledge, professional expertise, and international experience.

Khachatrian believes it is also important to increase the number of laboratories. One or a few small research groups cannot provide the scale of scientific activity needed for Armenia to become a prominent center for physical artificial intelligence.

He considers it important for the country to establish enough laboratories and research teams to consistently train specialists and develop innovations of interest to the international technology market.