Agriculture looks simple from the outside: open fields, slow speeds, predictable conditions. It isn't. Plants change season to season. Weather, dust, heat, and debris are constant. Terrain shifts across farms, geographies, and crop types. And seasonality creates real time pressure on every operation.

Whether the environment is an orchard, a construction site, or a mine, the core problems are the same: navigating without relying on GPS, identifying and responding to obstacles in real time, handling variable lighting and unstructured terrain, and generalizing without retraining from scratch.

In this session, we'll break down why agriculture is the right place to build physical AI first, and how that foundation can scale into other off-road industries.

Deploying rugged AI in specialty crops leaves no room for perception drift, timing jitter in control loops, or flaky edge connectivity. In this technical case study, Agtonomy CEO and Co-Founder Tim Bucher details how the team and their OEM partners engineer a factory-fit physical AI stack with zero-shot deployment, including real-time perception and planning, drive-by-wire integration, and middleware that abstracts heterogeneous OEM platforms and implements.

Attendees will dive into how this stack orchestrates multi-implement autonomy, handles failure modes in harsh environments, and maintains deterministic behavior and observability at the edge.

At Open Robotics, we believe that relying on tested, interchangeable software components facilitates the building of rugged autonomous systems. Using trusted building blocks from the Robot Operating System (ROS), developers can piece together smart AI applications for unpredictable farming environments.

This session explores how open-source software accelerates and improves the reliability of robot development by leveraging standard data types, field-proven data transport libraries, and shared data collection tools.

We'll also explore how recent additions to the ROS project, such as abstract zero-copy data buffers and ROS Bag utilities, can help move your data from the field to model training and back to deployed hardware.

The future of farming can't wait for the next growing season. This talk explores how NVIDIA's full Physical AI stack, from Cosmos, Omniverse and Isaac Sim to DGX/HGX, Jetson and IGX, helps agricultural innovators break through the industry's biggest bottleneck: scarce, seasonal and highly variable real-world data.

See how simulation, synthetic data and rugged edge AI can accelerate the development of smarter machines built to perceive, reason and act in one of the world's most unpredictable environments: the field.

No description provided.

The agricultural robotics industry has spent the last decade proving that autonomy is possible. From perception and navigation to machine learning and precision operations, innovators have demonstrated that autonomous systems can successfully operate in some of agriculture's most challenging environments. The next decade will be defined by a different challenge: industrializing autonomy.

The question is no longer whether autonomous systems can work. The question is whether they can be delivered, supported, and trusted at scale. This panel will explore what it takes to move from successful prototypes and pilot deployments to reliable, scalable, and commercially sustainable products. Panelists will discuss the realities of building businesses capable of deploying hundreds or thousands of autonomous systems while maintaining quality, reliability, safety, customer satisfaction, and healthy economics.

The discussion will also examine an important distinction within the industry. While "ag robotics" is often discussed as a single category, the challenges of scaling robotic implements can differ significantly from those of scaling autonomous mobility platforms. Neither is better than the other, but they represent different technical, operational, safety, and commercial problems to solve.

As the industry matures, success will increasingly depend not only on technical innovation, but also on operational excellence, product quality, customer support, safety standards, distribution strategy, and sustainable business models.

This session will bring together perspectives from startups, OEMs, investors, and growers, with a particular focus on what startups and OEMs must build in order to move from technical validation to durable commercial scale.

Drawing on multiple seasons of building and shipping AI-powered machines, Gabe Sibley walks through what it actually takes to close the loop from sensing to action in a real field, with real crops, under real conditions. The talk centers on Verdant Robotics™ and its SharpShooter™ system, which uses Aim & Apply™ technology to see, decide, and apply at the individual plant level in real time, and why that precision requires hardware and software co-designed from the start.

Gabe covers the architectural decisions that make physical AI work in agriculture: edge decision authority, hardware and software co-designed for worst-case field conditions from day one, confidence-gated execution, and human override as a built-in requirement rather than a fallback.

Attendees leave with a clear framework for what separates field-ready AI from open-loop tools that generate data but stop short of creating value.

The Diverging Paths of Product Development

As the Smart Machines and IoT Go To Market leader for North America at AWS, his primary focus is partnering with companies and solution providers to accelerate the deployment of Smart Machines and Physical AI solutions using the combined portfolio of AWS and partner services. His goal is to connect industrial assets to the AWS Edge and Cloud to collect physical intelligence and enable insights that drive real-world business outcomes such as operational efficiency, quality inspection automation, downtime reduction and sustainability.

The solutions we can help deliver range from initial real-time visualization to AI-driven task automation to fully autonomous machines across auto & manufacturing, energy and utilities, agriculture, construction and mining.

Rugged AI in Agriculture: What's Next. As artificial intelligence moves from pilot projects to real fields, the question isn't whether AI can work in agriculture—it's whether it can survive there. Dust, heat, unreliable connectivity, and the unforgiving economics of specialty crops demand AI systems that are as rugged as the environments they're built for.

This panel brings together perspectives from technology validation, grower-facing research, and enterprise AI scaling to explore what "rugged AI" really means in practice: how emerging technologies get tested and proven in the field, what growers actually need to trust and adopt automation, and how AI platforms move beyond pilots into systems that drive real P&L results. Panelists will discuss the practical barriers to AI adoption on the farm, what separates promising prototypes from market-ready solutions, and where the industry is headed as automation and AI become core infrastructure for specialty crop agriculture.

Panelists:

• Moderator Dan Kurdys (9 North Group)
• Ben Palone (Western Growers Innovation Team)
• Dr. William Turechek (California Strawberry Commission)
• Frank Faulring (Faulring Mechanical Devices, Inc)
• Matt Hoffman (Reservoir)
• Dr. Bo Liu (Cal Poly)