Rhoda AI is developing robotic foundation models that combine visual understanding, physical intelligence, and tactile sensing to help robots perform complex tasks across manufacturing and industrial environments.
Rhoda AI robot removing items from a box using learned manipulation skills.
Rhoda AI is developing a new approach to robotic intelligence that allows machines to learn physical behaviors through visual information. Rather than relying entirely on manually programmed movements, its technology uses video-based learning to help robots understand how objects move, interact, and respond to physical actions.
This approach is designed to make robotic systems more adaptable to changing environments and unfamiliar tasks. By learning patterns of physical interaction, robots could perform a wider range of operations without requiring a separate control program for every individual movement.
At the center of Rhoda AI's research is Direct Video-Action, or DVA, a model that connects predictions of future visual states with robotic actions. The system uses visual observations to anticipate how a scene should change and generate movements that help produce those changes.
This creates a framework for learning robotic behaviors from demonstrations and physical interactions. Rather than treating perception and movement as completely separate processes, DVA brings them together to support more flexible robotic manipulation in real-world environments.
Rhoda AI is developing advanced tactile sensors that allow robotic systems to detect and interpret physical contact with remarkable sensitivity. Its technology can measure forces as small as a few millinewtons while also handling forces exceeding 40 newtons, providing a wide operating range for manipulating objects of different sizes, weights, and materials.
Beyond measuring pressure, the sensors can detect directional forces, vibrations, and temperature changes. These capabilities provide robots with information about slipping, surface contact, and physical resistance, helping them make more precise adjustments when handling objects that require careful manipulation.
Rhoda AI is integrating tactile feedback into its Direct Video-Action framework, extending robotic learning beyond visual observations. By incorporating touch into the model's predictions, robots can learn how physical contact changes during manipulation and adjust their movements based on sensory feedback.
In its research demonstrations, Rhoda AI showed robots identifying and retrieving textured objects even when their cameras were obstructed. These experiments demonstrate how tactile intelligence can complement vision and potentially improve robotic performance in tasks such as precision assembly, object sorting, and handling delicate or irregularly shaped materials.
Rhoda AI is exploring how general-purpose robotic intelligence can support industrial operations that have traditionally been difficult to automate. Its demonstrations include tasks such as bearing decanting, container breakdown, and returns processing, where robots must interact with objects that vary in position, shape, and physical characteristics.
These applications highlight the potential for robotic systems that can adapt to changing workflows rather than relying exclusively on fixed movements and carefully controlled environments. In manufacturing and logistics, this flexibility could help businesses automate repetitive material-handling tasks while reducing the engineering effort required to accommodate different products and processes.
Combining video-based learning with tactile sensing could expand the range of tasks that industrial robots can perform reliably. Visual information helps robots identify objects and plan movements, while tactile feedback provides additional information about contact, grip stability, and physical resistance during manipulation.
For businesses evaluating automation, the long-term opportunity is to deploy robotic systems that can learn new tasks and operate across multiple workflows without extensive reprogramming. Although commercial performance will depend on factors such as reliability, throughput, integration costs, and maintenance requirements, Rhoda AI's research represents an important step toward more versatile robotic systems for industrial environments.
Explore ABB's industrial robots, collaborative automation, autonomous mobile robots, AI-powered applications, and advanced software for manufacturing and logistics.
Explore KNAPP's warehouse automation technologies, including robotic picking, automated palletizing, shuttle storage, autonomous mobile robots, sorting systems, and intelligent logistics software.
An examination of how the UK's Cooperative Technologists network connects worker-owned technology businesses, shares resources, and uses democratic governance to build an alternative model for the digital economy.
Exploring new models of democratic participation through sortition, rotating leadership, collective investment, and technology-enabled cooperative governance.
An exciting head-to-head competition where humanoid robots tackle real-world challenges, showcasing the latest advances in robotics, artificial intelligence, and autonomous technology.
Developing highly automated food production facilities that combine robotics, industry expertise, and regional partnerships to reduce costs and supply prepared food at commercial scale.
