About Haihe Laboratory of Brain-Computer Interaction and Human-Machine Integration

Haihe Laboratory of Brain-Computer Interaction and Human-Machine Integration

The Haihe Laboratory of Brain-Computer Interaction and Human-Machine Integration is a municipal research laboratory of Tianjin, in northern China, working on brain–computer interfaces (BCI) and the fusion of human and machine intelligence. Established in March 2023 and led by Tianjin University, it is the sixth Haihe Laboratory supported by the Tianjin municipal government. It works principally on non-invasive brain–computer interaction — reading brain signals without implants — and human–machine fusion, and Tianjin University reports that it holds the world's largest patent pool in brain–computer interaction. In English-language sources its name also appears as the Brain-Computer Interaction and Human-Machine Fusion Haihe Laboratory and as the Haihe Laboratory for Brain-Computer Interaction and Human-Computer Integration.

The sixth Haihe Laboratory

The Haihe Laboratories are Tianjin's provincial-level flagship system, whose first five laboratories — modern Chinese medicine, synthetic biology, cell ecology, materials green creation and manufacturing, and advanced computing and key software — were approved and unveiled in 2021. The brain–computer laboratory, established under Tianjin University's leadership in March 2023, extended the system into a sixth direction, brain–computer interaction and human–machine integration. Georgetown University's Center for Security and Emerging Technology (CSET), in a March 2024 bibliometric brief on China's non-therapeutic BCI research, described the laboratory in the same terms: established in March 2023 under Tianjin University's leadership, a focal point for a consortium of national institutions and some 60 top scientists, led by academician Gu Xiaosong of the university's School of Medicine, and aiming to solve the technical problems delaying BCI's wider adoption.

Leadership: Gu Xiaosong and the Tianjin team

The consortium is led by Gu Xiaosong, an academician of the Chinese Academy of Engineering at Tianjin University's School of Medicine, known for his work in neural regeneration and tissue engineering. The university's BCI research is publicly fronted by its neural engineering team; Ming Dong, a vice-president of Tianjin University, serves as an executive director of the laboratory and has been its principal spokesman for its brain-on-chip work, and researcher Xu Minpeng has represented the team in its adaptive-interface research. CSET's count of some 60 top scientists in the consortium indicates the laboratory's model: like the other Haihe laboratories, it assembles an existing national-strength team — Tianjin University describes its BCI group as one of China's earliest — into a municipal platform with a defined mission.

The 216-command world record

The laboratory's best-documented early result came on 18 May 2023, at the 7th World Intelligence Congress in Tianjin, where it officially released an independently developed non-invasive brain–computer interface system with a very large command set and high speed. Tianjin University's report records that a wearer looking at a virtual keyboard of 216 keys could spell and communicate by mind-typing: the system set a world record for the largest instruction set of a non-invasive brain–computer interface, with online average information transfer above 300 bits per minute and an average single-command output time of 1.2 seconds, and the technology was granted patents in China and the United States. The university explained the method: a time–frequency–phase hybrid multiple-access coding paradigm integrating three EEG characteristics — movement-induced potential, P300 potential and steady-state visual evoked potential. It also placed the result in the team's own trajectory: first to a hundred-instruction system in 2019, then past the 200-instruction mark, then to 216.

The patent pool

Tianjin University states that the laboratory has laid out the world's largest and most comprehensive patent pool for brain–computer interaction, and that it holds the key technology of non-invasive brain–machine interaction. Patent position matters in BCI because the field is moving from laboratory demonstration toward medical devices: the laboratory's later reporting records that its team has assisted in bringing more than 20 BCI medical devices to market, benefiting thousands of patients. The laboratory also jointly confers, with Tianjin's National Key Laboratory of Advanced Medical Materials and Devices, the Huanao Award for BCI research — a sign of its position as a national convening point for the field, documented in CSET's brief through the award's 2023 recipients.

MetaBOC: brain-on-chip interaction

In 2024 the laboratory's team, with the Southern University of Science and Technology, presented MetaBOC, described as the world's first open-source brain-on-chip intelligent complex information interaction system. The system connects a brain organoid — brain tissue cultured in vitro from stem cells, with some of the intelligent functions of a biological brain — through an electrode chip to external devices, enabling a robot to perform tasks such as obstacle avoidance, tracking and grasping. Tianjin University's Ming Dong explained its two departures from earlier work: cell culture moving from two dimensions to three, providing a more complex neural computing network for the brain-on-chip, and the integration of artificial-intelligence algorithms, enabling attempts at hybrid intelligence. The study was reported in the journal Brain, including work on low-frequency ultrasound stimulation of organoid neural networks. Brain-on-chip, Ming Dong said, is emerging as an important new branch of brain–computer interfaces, distinct from technologies that use a living human or animal brain as the experimental subject.

Two-way adaptive interfaces

Researchers from Tianjin University, with Tsinghua University colleagues, also developed what they described as the world's first two-way adaptive brain–computer interface, published in Nature Electronics in 2025. Conventional BCIs decode in one direction only; the new system adds feedback so that brain and machine learn from each other — a concept the Tianjin co-author Xu Minpeng called brain–computer co-evolution. Its decoder is built on a memristor chip, whose neural-network-like architecture is energy-efficient, in a dual-loop design: a machine-learning loop updates the decoder as brain signals change, and a brain-learning loop gives the user feedback to refine control. The team reported efficiency gains of about 100-fold and energy-demand reductions of about 1,000-fold against conventional devices, accuracy improvements of 20 per cent, stability over six-hour sessions with ten participants, and four degrees of freedom in brain-controlled drone flight. The work illustrates the laboratory's wider Tianjin research environment at its most ambitious: interfaces that adapt to the brain, not only the reverse.

Miniaturisation: the 3-gram system

The laboratory's drive toward practical devices is also documented in miniaturisation. In 2026 Tianjin University reporting, carried in technology press, the laboratory unveiled a fully integrated non-invasive BCI system of about 3 grams in roughly 2 cubic centimetres — described as the smallest and lightest such system to date — combining EEG electrodes, signal circuitry, a battery and wireless transmission in a single unit small enough to be hidden in hair. Its reported sampling rate is 1,000 Hz, with eight to ten hours of operation per charge, and its primary metrics were stated to meet China's national medical-device standard for electroencephalographs, GB 9706.226-2021. Identified applications included clinical monitoring, brain-controlled games, multi-person hyperscanning and workplace safety. The laboratory presents its scope as spanning the whole technology stack: chips and electrodes, core algorithms, system integration and platforms.

Why non-invasive matters

The laboratory's consistent emphasis on non-invasive interfaces defines its position in world BCI research. Invasive systems, implanted in the brain, offer high signal quality at the cost of surgery; non-invasive systems, reading EEG from the scalp, are safer and cheaper but must extract commands from far weaker signals — which is why command-set size, transfer rate and decoding algorithms, the laboratory's documented strengths, are the field's contested metrics. The 216-command record, the adaptive decoder and the 3-gram integrated device are three answers to the same problem: making non-invasive BCI fast, stable and wearable enough for communication aids, rehabilitation and everyday human–machine fusion — the laboratory's stated aim of solving the technical problems delaying BCI's wider adoption.

Significance

The Haihe Laboratory of Brain-Computer Interaction and Human-Machine Integration is the most publicly visible of Tianjin's Haihe laboratories, because its results arrive as demonstrations the public can picture: typing 216 commands by thought, a cultured brain organoid steering a robot, a drone flown by an adaptive interface. Its documented foundations are correspondingly concrete — establishment in March 2023 as the sixth Haihe laboratory, leadership by Tianjin University and academician Gu Xiaosong, a consortium of some 60 leading scientists recorded by an independent American research centre, a world-record non-invasive command set, the field's largest patent pool by its university's account, and a line of work from MetaBOC to two-way adaptive interfaces. It is Tianjin's municipal platform for the point where neuroscience, engineering and artificial intelligence converge.