ADI at NVIDIA GTC 2026: Showcasing Physical Intelligence in Action

NVIDIA GTC 2026 returns to San Jose on March 16, and Analog Devices, Inc. (ADI) is participating to demonstrate how physical intelligence is transforming robotics.

This year, we’re leaning all the way into physical intelligence: the convergence of sensing, signal processing, actuation, and AI that enables robots to operate with humanlike capability. The focus of ADI’s presence is showcasing the AI-ready humanoid hand platform: ADI’s motor drive and sensing architecture designed to unlock true robotic dexterity.

Dexterity remains a major bottleneck in automation. While humans effortlessly manipulate objects, robots continue to struggle with tasks that require fine motor control adaptability and touch. Progress is further slowed by the lack of common metrics, agreed-upon difficulty levels, or reproducible test conditions. To help move the industry forward, ADI will preview a new industrial dexterity benchmark created to objectively assess and accelerate progress in robotic manipulation. Grounded in physics-aware sensing, high-fidelity simulation, and rigorous evaluation, ADI aims to provide the industry with a shared yardstick that accelerates innovation, validates performance claims, and enables meaningful comparisons across platforms.

Visitors to our space, Booth 3203 in the Grand Ballroom, will experience ADI’s vision for robotic dexterity through three immersive demonstrations.

Demo 1: Tactile Intelligence — Multimodal Sensing for Dexterous Manipulation

What if a robot could feel its way through a task?

This demo highlights ADI’s next generation‑ multimodal tactile sensor prototypes, capable of detecting contact, slip, and fine surface interactions in real time. Using tactile feedback alone (no vision), the humanoid hand traces along a network cable and finds its way to a socket. The model behind the demo was developed using NVIDIA Isaac Sim™, where ADI implemented a digital twin of the tactile sensor, to accelerate sim-to-real policy transfer.

Why it matters: Vision alone can’t solve dexterity. With ADI’s tactile sensors, algorithms, and physical intelligence stack, robots gain the nuanced contact awareness required for next-generation humanoid platforms and generalized manipulation tasks. But how do you measure real‑world performance? That question leads directly to Demo 2.

Demo 2: Industrial Dexterity Benchmark & Digital Twin

Working in collaboration with internal and external stakeholders, ADI is building on National Institute of Standards and Technology (NIST) dexterity examples to develop a new set of application-grounded dexterity benchmarks. Initial use cases include:

Data center cable manipulation
Automotive cable harness routing and installation
Precision planetary gearbox assembly

To address the sim-to-real challenges inherent in these tasks, ADI is partnering with Synopsys to develop accurate and fast collision and soft-body simulation capabilities.

Comments

Join the conversation at EngineerZone

Opinions expressed in the content posted here are the personal opinions of the original authors, and do not necessarily reflect those of Analog Devices, Inc (ADI) or its subsidiaries. The content is provided for informational purposes only and is not meant to be an endorsement or representation by ADI or any other party. This site may also provide links or references to non-ADI sites and resources. ADI makes no representations, warranties, or other commitments whatsoever about any non-ADI sites or third-party resources that may be referenced, accessible from, or linked to this site.

This demo introduces the benchmark through an interactive digital twin of the AI-ready humanoid hand platform. The simulation - running in NVIDIA Isaac Sim on an NVIDIA RTX PRO^TM6000 GPU - shows the hand attached to a robotic arm and working with a high-fidelity cable. This first benchmark test case demonstrates ADI’s commitment to realism and repeatability.

Customers can use the environment to generate synthetic data, train and validate manipulation policies, and achieve more reliable sim‑to‑real transfer—reducing physical prototyping costs and accelerating product development.

Demo 3: AI-Driven Humanoid Hand Cable Manipulation

If demo #1 is about “touch” and demo #2 is about the development process, then demo #3 brings everything together.

This live demonstration showcases end-to‑end manipulation: sensing, perception, control, and execution working as a single system. A humanoid robotic hand mounted on a robot arm identifies a cable, grasps it, manoeuvres it precisely, and plugs it into a network socket approaching human ‑level fine motor capability.

ADI’s engineers developed and validated the system using NVIDIA Isaac Sim, with NVIDIA SkillGen supporting the generation and refinement of manipulation skills. The project also gives ADI ongoing first-hand insight into customer workflows and informs future product enhancements.

Plus: Bi-Manual Robot Dexterity Demo at the Synopsys Booth

ADI’s work on robotic dexterity continues at Synopsys Booth #1135 in Exhibit Hall 2.

This joint demo integrates ADI’s next generation Tactile Sensors with Time-of-Flight sensing and Ansys Mechanical multiphysics simulation to enable advanced intelligent cable manipulation. It combines an NVIDIA Jetson Thor^TM powered bi-manual robotic arm setup from Flexiv with force, vision, and contact sensing, all connected within a high-fidelity digital twin.

The result is predictive, optimized, and robust manipulation performance in dynamic industrial environments - showcasing what’s possible when hardware, simulation, and AI converge.

ADI Talks at GTC 2026

ADI’s team is participating in several panels and talks throughout NVIDIA GTC.

Paul Golding, Vice President of Edge AI, will participate in the panel session ‘The Edge of the Edge: Redefining GPU-Enabled AI Sensor Processing', which takes place Wednesday, March 18 at 3:00 p.m.
Philip Sharos, Principal Engineer at ADI’s Edge AI team will co-host Synopsys’ Expo Talk ‘Reducing the Sim2Real Gap for Industrial Robotics’ on Monday, March 16 at 3:00 p.m. This session, presented by Synopsys, discusses the critical need for new sensing modalities, and high‑fidelity simulation techniques that accelerate learning and significantly narrow the Sim‑to‑Real gap for industrial manipulation. The presentation will also touch on ADI’s work on novel dexterity benchmarks and Synopsys contributions to deformable objects and connector assets.
Tao Yu, Director, Machine Learning at ADI’s Emerging Business Unit will participate in a booth talk at the Lightwheel booth, Booth 1406, on Tuesday. March 17 at 2pm. This talk presents a joint effort between Analog Devices and Lightwheel to improve simulation fidelity for high‑precision robotic tasks. The work focuses on a real2sim validation pipeline that compares measured forces, contact behavior, and deformable interactions against simulated results. By pairing advanced sensing with a calibrated physics engine, the collaboration provides a clear way to evaluate where simulation aligns with physical measurements and where it does not. The goal is to support more dependable training and evaluation for manipulation systems that depend on subtle physical feedback.

From tactile intelligence and humanoid manipulation to industrial benchmarks and digital twins, ADI is showing what’s possible with physical intelligence.

If you’re building next‑generation robots, exploring humanoid platforms, or scaling autonomous industrial systems, visit us in Booth #3203 at NVIDIA GTC 2026. We look forward to connecting with the robotics community and our partners - and sharing what’s next.