David Lin | GE Aerospace Research: Why does a world-class MEMS gyroscope lose 25x its performance the moment you package it?

How can we bypass the latency and power bottlenecks of traditional silicon architectures in tactile robotics?

The integration of physically embodied neural networks directly at the sensing node represents a paradigm shift in edge computing. By executing complex operations such as edge detection and real-time image sharpening locally, these systems eliminate the need to transmit raw data to centralized computer vision processors. This localized processing architecture bypasses the high power consumption and bandwidth constraints of conventional silicon-based computing, enabling immediate signal conditioning at the point of data collection.

In high-precision robotics, communication latency is a critical failure point; a delay of 100 milliseconds during complex tactile manipulation is unacceptable. Utilizing flexible hybrid electronics (FHE) allows for localized, single-digit millisecond feedback loops directly at the physical interface. Rather than treating intelligence as an external processing unit latched onto a mechanical frame, FHE enables true morphological intelligence, where the structural body of the robot itself is inherently smart, responsive, and computationally active.

Achieving this level of structural integration requires moving away from discrete component assembly toward multi-material 3D printing. By continuously integrating sensors, actuators, and control circuitry, we can modulate material properties to measure entire physical fields directly, such as localized stress and friction, rather than relying on sparse, point-wise data. This biomimetic approach distributes sensing, actuation, and computation throughout the physical structure, leveraging material physics to overcome the non-linear scaling limits of Moore's law.

In this short video, you can learn:
* How physically embodied neural networks perform real-time edge detection and image sharpening at the sensor node.
* The critical role of single-digit millisecond latency in tactile robotics and how FHE enables morphological intelligence.
* How multi-material 3D printing transitions robotics from point-wise data collection to direct physical field measurement.

📋 **Clip Abstract** The speaker discusses how flexible hybrid electronics can embed real-time edge detection, image sharpening, and low-latency processing directly into a robot's physical structure. By distributing sensing, actuation, and computation biomimetically, this approach overcomes the scaling and latency limits of centralized silicon-based computing.

🎤 Speaker: David Lin
🏢 Company: GE Aerospace Research
📅 Event: Future of Electronics RESHAPED USA 2026
📍 Location: Computer History Museum, Mountain View, California, USA

🌐 Learn more at the next TechBlick event: https://www.techblick.com

How can we bypass the physical bottlenecks of centralized processing in next-generation soft robotics?

The evolution of intelligent robotics is reaching a threshold where algorithmic optimization alone cannot overcome physical latency and wiring complexity. To bridge this gap, next-generation robotic systems must transition toward physical intelligence, embedding sensing, actuation, and processing directly into the structural material. Utilizing Flexible Hybrid Electronics (FHE) allows designers to mimic biological systems, distributing functional intelligence across the physical substrate rather than relying on a distant, centralized processor.

While high-density sensory arrays are mature, routing massive volumes of raw sensor data to a centralized ASIC creates severe communication bottlenecks. In biological organisms, localized peripheral processing filters out environmental stimuli before signals reach the brain. Replicating this efficiency in soft robotics requires shifting away from centralized architectures toward localized edge computation, reducing the bandwidth demanded of the primary processor.

To achieve this decentralized compute distribution, researchers leverage reservoir computing to exploit the non-linear physical dynamics of the robotic structure. By utilizing the material's inherent properties to perform localized mathematical transformations, a reservoir computing network compresses massive streams of raw sensory data. This localized preprocessing filters out noise, allowing the system to transmit only critical, high-level features to the central ASIC and reducing communication overhead.

In this short video, you can learn:
* Why algorithmic advancements alone are hitting a limit and how FHE enables the next generation of embodied robotic intelligence.
* The architectural limitations of routing high-density sensor data to a centralized ASIC.
* How reservoir computing exploits structural dynamics to compress raw data and decentralize computation in soft robotics.

📋 **Clip Abstract** The speaker discusses how embedding Flexible Hybrid Electronics (FHE) directly into robotic structures can overcome the limits of traditional algorithms by distributing intelligence into the material itself. He explains how reservoir computing can resolve the communication bottlenecks of centralized ASIC architectures by processing and compressing raw sensory data locally within the robot's physical structure.

🎤 Speaker: David Lin
🏢 Company: GE Aerospace Research
📅 Event: Future of Electronics RESHAPED USA 2026
📍 Location: Computer History Museum, Mountain View, California, USA

🌐 Learn more at the next TechBlick event: https://www.techblick.com