Home NeuroXess Unveils 'SimLife' Trial: Fully Implanted Brain-Computer Interface Enables Paralyzed Patients to Perform Continuous Real-World Tasks with 84% Success Rate

NeuroXess Unveils 'SimLife' Trial: Fully Implanted Brain-Computer Interface Enables Paralyzed Patients to Perform Continuous Real-World Tasks with 84% Success Rate

Nov 17, 2026 14:36 CST Updated 14:36
NeuroXess

Invasive Brain-Computer Interface Developer

Shanghai’s leading brain-computer interface (BCI) company, NeuroXess, recently disclosed its latest explorations based on its independently developed fully implanted, fully wireless, and fully functional (“three-full”) BCI system. Within its “SimLife” future testing arena, clinical participants conducted continuous task validations across four real-life scenarios: independent mobility, home living, social interaction, and work.

Execute continuous living tasks

In this exploratory study, the clinical participant, Mr. Song (a pseudonym), suffered high-level paraplegia due to a car accident, resulting in an inability to voluntarily move below the elbows. Within the “SimLife” Future Testbed, NeuroXess established a complete technological pipeline spanning from intention decoding and intelligent perceptual decision-making to coordinated device execution and feedback. Notably, the core latency of the system—from receiving neural signals to completing decoding and outputting control commands—was less than 50 milliseconds, effectively supporting the participant’s transition from simple screen-based operations to performing coherent activities of daily living.

In this workflow, the “Three-Comprehensive” brain-computer interface (BCI) is responsible for precisely decoding intentions emitted by the brain, while large AI models and intelligent vision systems take on the critical tasks of environmental understanding and decision support. In scenarios involving independent travel and shopping, participants transmit movement intentions via the BCI; the perception system real-time identifies traffic lights, gate barriers, and shelf statuses to assist electric wheelchairs in obstacle avoidance and navigation, while simultaneously coordinating with exoskeleton gloves to enable precise grasping and brain-controlled payment settlement. In home living environments, the AI system perceives objects of attention through eye-tracking and interfaces with smart home (IoT) devices; after brain-controlled ordering is completed, the intelligent hub orchestrates service delivery.RobotDelivered, ultimately assisting with independent eating through a brain-controlled exoskeleton glove. In independent work and social interactions, participants can not only smoothly control devices in their workspace but also chat online with friends, make video calls, and even engage in remote chess matches and other interactive activities.

The successful execution of these sequential actions is attributed to the synergy between intent recognition and device actuation. In multi-device switching tasks, the system’s intent recognition accuracy approaches 100%; participants achieved a stable success rate of over 95% in grasping small objects, such as chess pieces, using brain-controlled exoskeleton gloves. In continuous tasks involving going outdoors, obstacle avoidance, and forward movement, clinical participants achieved an 84% success rate in completing the entire process seamlessly in a single attempt, demonstrating the system’s coordinated capabilities from intent recognition and device operation to the execution of continuous tasks.

Efficiently Understand Complex Environments

Continuous living task validation demonstrated the system’s interactive capabilities, while constructing a model of the real physical world represents NeuroXess’s higher-level strategic direction. As brain-computer interfaces move from screens into the real physical world, the core challenge for the system has escalated from the precise execution of single actions to the efficient understanding of dynamic and complex environments.

In the real physical world, the environment is full of variables. Through continuous interaction in the "The Sims" future testbed, the system needs to constantly understand: What kind of space is the user in? What stage has the current task reached? What sudden changes have occurred in the environment and equipment? Only by comprehending these contexts can machines truly transition from passive execution to active understanding.

Through this continuous cross-scenario interaction, the system gradually establishes deep correlations among humans, intents, environments, devices, and tasks. This opens up new research directions for underlying algorithms and models: rather than relying on massive repetitive training centered around fixed actions (such as grasping or clicking), the system explores generalization capabilities across tasks and scenarios. As the system accumulates extensive action feedback from real-world environments, it can transfer these experiences to entirely new tasks. This is the core objective of NeuroXess in advancing real-world physical world models.

This experience transfer is also reflected in cross-scenario applications: after training in specific scenarios, when the system transfers home control experience to new scenarios such as offices, intent recognition and task adaptation capabilities are significantly improved. Specifically, for learning new tasks, supported by contextual understanding capabilities of real-world physical models, participants' average training time to master new continuous tasks like shopping has been reduced from 10.5 hours to 2 hours, providing practical evidence for exploring the transfer and reuse of real-world scenario experiences.

Patient Benefit as the Benchmark

Authentic, multi-task interactions not only provide fertile ground for model iteration but also lay the foundation for ecological synergy connecting large language models, embodied devices, and intelligent spaces. This testbed can simultaneously accommodate 20 individuals for training, with an estimated annual accumulation of approximately 18,000 hours of interaction data, providing continuous support for the training and iteration of real-world physical models.

However, whether in model construction or ecological synergy, the ultimate criterion for validation remains the tangible benefits to patients. Whether a procedure is successful or a specific metric improves merely represents dimensions for observing system capabilities; what truly determines how technology transforms actual life is whether patients can more independently perform daily activities, engage in work, travel, and participate in social interactions. In the training of real-world physical models, the duration of independent activity per day without caregiver assistance increased by 51%, which serves as a concrete manifestation of this benchmark.

Reporters learned that the "SimLife" Future Testbed brings forward issues from future applications into the present, allowing patients to explore the application of their capabilities here and lay the foundation for more complex applications later. These continuously accumulated real-world interaction data will ultimately feed back into the scenario and system iterations of the "SimLife" testbed. This closed loop, originating from reality and applied to reality, is ensuring that each training session better connects with patients' real lives. By our reporter, Gao Yang.