Robotics moves one step closer to precise integration with complex environments thanks to the DexNDM model, a solution that redefines the paradigm of dexterous manipulation in robots. This model, designed to simulate neural dynamics in robotic manipulators, allows objects to be rotated and adjusted with unprecedented precision, encompassing small, elongated shapes and irregular geometries.
Robotic skills in simulation and reality
DexNDM overcomes one of the main barriers in modern robotics: the effective transfer of simulations to the real world. Thanks to its neural architecture, a single model can adapt to multiple wrist orientations and environmental conditions without requiring manual adjustments, optimizing deployment in highly dynamic tasks.
The model has been trained to perform complex manipulations directly in the palm of the robot. This includes intricate rotations, internal twists and full coverage of the object during execution, which improves performance in operations where the margin of error must be minimal.
Operational robustness in long-haul tasks
Designed for both simulations and real physical scenarios, DexNDM maintains a high level of performance even under prolonged tasks or varying conditions. This robustness makes it an ideal tool for applications requiring continuous adaptability and sustained accuracy over time.
With extended capabilities to handle complex objects, DexNDM is emerging as a key resource in sectors such as advanced manufacturing, fine assembly, automated inspection and assistive robotics. Its flexibility allows it to operate without relying on specific geometries, representing a clear competitive advantage over more traditional models.
The evolution of this technology continues to integrate with advancements in cognitive robotics, allowing intelligent systems to not only execute commands but to understand and adjust their manipulation according to the physical and dynamic context of each task.
Source: DexNDM
Photo: Freepik