Scientists have created a "neurohelmet" to control a robot dog - ForkLog: cryptocurrencies, AI, singularity, the future

# Scientists have created a “neurohelmet” to control a robot dog

At Xi’an Jiaotong University, a team led by Professor Xu Guanghua has developed a way to control a robot dog using nothing but thought.

Mind-controlled robot dogs? They’re real now at China’s Xi’an Jiaotong University. Just think “forward” — and the robot dog moves. Professor Xu Guanghua’s team made it happen with the non-invasive brain-computer interface technology. #FutureChina pic.twitter.com/3pIL7FffRv

— China Xinhua Sci-Tech (@XHscitech) March 29, 2026

The solution is based on non-invasive brain-computer interface (BCI) technology. It captures signals from neuronal activity using specialized sensors and enables precise control of mechanical devices.

Xu described the system as a kind of “remote control in the mind.”

When the user forms an intention such as “move forward,” the brain generates corresponding EEG signals. The system collects and decodes them, identifies the intended command, converts it into an instruction for control, and sends it to the robot dog, which carries out the task.

At the current stage, the solution supports 11 main commands such as moving forward, moving backward, and turning. The recognition accuracy exceeds 95%, and the delay between thought and action is about one second.

Invasive BCI technologies deliver high accuracy, but they require surgical implantation into the brain. This entails certain risks, including injuries, infections, immune rejection, and signal quality degradation over time.

Guanhua’s approach is considered safer, more cost-effective, and more convenient for users. However, non-invasive signals are less accurate, making it difficult to enable continuous, fine-grained real-time control.

The team decided to apply a hybrid approach that uses a human-machine collaboration model.

“People are only responsible for formulating high-level intentions like ‘where to go.’ At the same time, highly accurate, high-speed, and repetitive tasks—such as autonomous navigation, perceiving the environment, dynamically avoiding obstacles, and performing movements—are carried out by the machines’ own intelligent systems,” — said Guanhuа.

This approach improves the efficiency and stability of the solution, helping to bypass limitations related to the accuracy of non-invasive signals.

The scientist noted that for BCI to advance, regular breakthroughs are needed in core technologies, along with deep integration with cutting-edge fields such as artificial intelligence and autonomous navigation.

According to Guanhuа, the robot dog could become a good assistant for people with disabilities.

Recall that from March 25 to 29, the ZGC Forum was held in Beijing, where developers showcased a wide range of products in the BCI field—from specialized chips to rehabilitation systems.