Agility Robotics' Digit Showcases Autonomous 'Shopping' in New Demo

Agility Robotics' Digit Showcases Autonomous "Shopping" in New Demo

Agility Robotics has released a new video demonstrating its humanoid robot, Digit, performing autonomous mobile manipulation tasks in a simulated retail or warehouse environment. The video, published on YouTube on June 4, 2024, shows Digit navigating a space, picking up a red shopping basket, and then selecting various items from shelves to place into the basket.

Under the Hood: Simulation-Trained Control

According to Agility Robotics, the demonstration highlights Digit's capabilities driven by a "whole-body controller for mobile manipulation." A key aspect of this controller is that it was "trained in simulation." This sim-to-real approach is increasingly common in robotics, aiming to accelerate development and improve a robot's ability to adapt to real-world scenarios. The company states this training enables Digit to "execute tasks while navigating new environments and manipulating objects it has never encountered before."

Watch the new video below:

The "Shopping" Task

In the demonstration, Digit is seen first identifying and grasping a red shopping basket. It then moves to a set of shelves, where it picks up groceries and places them into the basket. While the setting is clearly a controlled laboratory environment with neatly arranged shelves and clear pathways, the fluidity of motion and the sequence of actions are notable. The ability to handle objects of different shapes and sizes is a critical skill for robots intended for logistics or retail applications.

Context and Implications

This demonstration aligns with Agility Robotics' stated mission to create "robots that are made for work," specifically targeting tasks that are "tedious and repetitive" for humans. The focus on mobile manipulation—combining locomotion with the ability to interact with objects—is crucial for humanoids designed to operate in human-centric spaces like warehouses, retail stores, or potentially even homes. If Digit can reliably generalize its learning from simulation to handle genuinely novel objects and navigate unfamiliar (though structured) environments, it would represent a significant step forward for practical humanoid applications.

Lingering Questions

While visually impressive, the video provides a curated glimpse into Digit's capabilities. The precise level of autonomy—for instance, whether object locations were pre-defined or if the robot is truly identifying and selecting items based on more general commands—remains an area for further clarification beyond the video's description. The definition of "new environments" and "objects it has never encountered before" also warrants more detailed explanation to fully assess the robustness and adaptability of the system. The transition from controlled lab settings to the complexities and unpredictability of real-world commercial environments continues to be a major hurdle for the robotics industry.

Looking Ahead

Agility Robotics' latest demonstration offers another data point in the accelerating development of humanoid robots. The emphasis on simulation-trained, whole-body control for autonomous tasks like item picking is a promising direction. As with all such advancements, the true test will be the robot's performance and reliability in less structured, real-world deployments.