Walking through the halls of The International Robot Exhibition (iREX 2025) at Tokyo Big Sight, the future felt tangible: robots navigating crowds, humanoids performing nuanced tasks, and service robots in roles that would have been beyond imagination a decade ago. iREX 2025 marks a shift in global robotics, as robots become partners in everyday environments, shaping the way we live, work, and move.

Humanoids gained some of the strongest public attention, especially models capable of flexible manipulation, adaptive motion, and semi-autonomous decision making. Companies like Flexiv Robotics showed robots performing tasks ranging from automotive body polishing to delicate assembly to personalized massage, all driven by sensor-rich “physical-AI” control systems. These systems signal a shift from pre-programmed rigidity toward robots that adjust dynamically to real-world conditions.

For that, humanoids will inevitably require distributed sensor intelligence.

NASP low-power, high-speed processors could serve as “the nervous system” for humanoids with a GPU-based brain. Like a human brain receiving sensory embeddings, the humanoid brain will receive data embeddings from NASP cores rather than raw data streams, saving power and time.

For IMU, tactile, audio, vibration, and force sensors, no existing signal processing solution offers microwatt power consumption and microsecond-level processing latency as POLYN’s NASP solution does. NASP could become the default sensor edge AI layer for the entire humanoid robotics sector.

Our meetings at iREX 2025 saw many discussions about the motion perception needed to improve robots’ reaction times. It is clear that NASP solutions are uniquely positioned to provide the necessary latency for balance, grip control, slip detection, proprioception, and collision avoidance.

Other advantages are power consumption and cost. Equipping every sensor with a Jetson or even a Cortex-M processor is hardly affordable. But a NASP NeuroSense core inside every sensor module is inexpensive and practical.