🤖 About the Project

The "Robot" project aims to create a unique ecosystem of intelligent robots capable of performing a wide range of industrial and domestic tasks. We plan to purchase several robot prototypes from different countries, train them using advanced artificial intelligence technologies, and adapt them for real-world applications.

📦 Which robots we plan to acquire

🇪🇺 NEURA 4NE1 (Germany)

A full-sized cognitive humanoid robot developed in collaboration with Studio F.A. Porsche. Cost — €98,000 per unit (up to 20 units).
Specifications: height 180 cm, lifting capacity up to 110 kg, speed 5 km/h, NVIDIA Thor T5000 processor. Expected to go on sale in late 2026.

🇨🇳 Agibot A2/G2 (China)

At MWC 2026, Agibot presented a full lineup of robots for various tasks.
Available models:

• A2 Series — full-sized humanoids for greeting guests and navigation
• G2 Series — wheeled robots for industrial tasks (precision assembly)
• D1 — quadruped robots for logistics and patrolling

Terms: available for purchase or rental starting from €900/day via the RaaS platform.

🇪🇺 NEURA 4NE1 Mini (Germany)

A compact version of a cognitive robot, created as a "Western answer" to affordable Chinese models. The cost ranges from €19,999 to €29,999 depending on the version.
Specifications: height 132 cm, weight 36 kg, 25 degrees of freedom, payload 3 kg, operating time ~2.5 hours.
Pro Version is equipped with 12-DOF arms and access to a digital twin. Expected to be on sale from April 2026.

🇨🌍0 UWORLD by UBTECH (China)

The world's first full-size hyper-bionic humanoid robot designed for emotional interaction.
Specifications:

• Male version: height 183 cm, weight 42 kg
• Female version: height 168 cm, weight 35.2 kg
• 88 degrees of freedom, 2-4 hours of operation

Price: pre-order with a deposit of 3,000 yuan (~30,000 ₽). Delivery starting September 2026.

🌍1 How we will train robots using AI

Robot training is a key part of our project. We will use the most advanced methodologies:

• Imitation Learning: Robots will learn by observing human actions. This is the fastest way to teach a robot new skills. For example, an operator can show the robot how to assemble a specific part once, and the robot will be able to repeat that action independently.
• Reinforcement Learning: Artificial intelligence will learn through trial and error, receiving a "reward" for correctly performed actions. This technology already enables groups of robots to work in harmony.
• Tactile Skills: We will use a breakthrough methodology described in the journal Nature Machine Intelligence, which allows robots to perform complex tactile tasks with nearly 100% success, even under changing conditions.
• Skill Generalization: A robot trained on a single task will be able to adapt to similar tasks without reprogramming. For example, after learning to grasp one type of part, it will be able to work with other shapes and sizes.

🌍3 Market Potential and ROI

The humanoid robot market is transitioning from concepts to mass production. Наши преимущества:

• Доступность по цене: Появляются модели по цене смартфона (от 110 000 ₽), что открывает массовый рынок.
• Модель «Робот-как-услуга» (RaaS): Возможность аренды от €900 в день позволяет тестировать решения без крупных вложений.
• Окупаемость: Сокращение времени на переналадку с недель до часов — экономия сотен тысяч рублей на каждом рабочем месте.

🌍4 Кто нам нужен