Open-Source Humanoid Robot Asimov 1 Tests New Grippers, Taking a Key Step Toward Grasping and Manipulation

From "Lifting" to "Picking Up": The Grasping Evolution of Asimov 1

The open-source humanoid robot project Asimov 1 has reached its 363rd day of development. According to the project author's share on Reddit, the team is testing brand-new grippers for the robot, with the goal of enabling it not only to lift objects but to truly grasp and interact with them.

Previously, Asimov 1's arms already had sufficient load capacity, but the lack of an end effector meant it could only perform "pushing" or "lifting" motions. The addition of grippers fills in a key missing piece of manipulation capability—shifting from simple force output to fine-grained grasping and handling. The author noted, "Now is the time to put the arms to use for manipulation."

Fully Open Source + Off-the-Shelf Parts: A Humanoid Robot Anyone Can Replicate

Asimov 1's most notable feature is that it is fully open source and built using off-the-shelf parts. The project code is hosted on GitHub (menloresearch/asimov-1), and anyone interested can build their own unit by following the drawings and bill of materials.

This approach stands in sharp contrast to most humanoid robots on the market. The latter often rely on custom actuators and closed supply chains, with unit costs running into tens or even hundreds of thousands of dollars; Asimov 1, by contrast, has chosen a "reproducible" path that lowers the hardware barrier and allows the community to iterate and improve around the same platform. The gripper design will likely follow the same logic—using easily sourced components to achieve usable grasping functionality.

Why Grasping Capability Is a Watershed for Humanoid Robots

In the field of humanoid robotics, walking and balance are often regarded as the first threshold, while manipulation is the second—and harder—threshold to cross. The reasons are:

• Complexity of the end effector: A gripper must simultaneously balance grasping force, precision, and adaptability, and different object shapes impose different requirements on finger structure;
• Coupling of perception and control: Grasping requires the coordination of visual localization, force feedback, and motion planning—far more than simply "closing the fingers";
• Cost and maintainability: Open-source projects must trade off between performance and manufacturability, and overly precise grippers would deter individual builders.

Asimov 1's choice to test grippers at this stage indicates that its basic locomotion and arm structure are already relatively stable, and the project's focus is shifting from "standing and walking" to "being usable."

End Effector is the component at the end of a robot arm that directly interacts with the outside world, most commonly in the form of grippers, suction cups, or dexterous hands. Its performance is typically measured along three dimensions: grasping force (how heavy an object it can lift), precision (whether it can align with millimeter-level features), and adaptability (whether it can stably grasp objects of different shapes and materials). Industrial settings commonly use two-finger parallel grippers, which are simple in structure and low in cost but prone to failure with irregular objects; multi-finger dexterous hands offer greater adaptability but bring higher actuation complexity and control difficulty. For humanoid robots, the end effector must also balance weight and volume, because the arm's load capacity is limited, and an overly heavy gripper directly reduces the upper limit of the weight of objects that can be carried.

Force Feedback / Force Control is the core technology that distinguishes grasping from lifting. Lifting an object only requires the arm to output sufficient torque, whereas grasping requires sensing the contact force between the fingertips and the object and adjusting in real time: too little force and the object slips, too much force and fragile items are crushed. A common approach is to integrate force sensors or current estimation at the gripper fingertips, using the measured contact force as closed-loop feedback, combined with impedance control to give the gripper spring-like compliance. Visual localization is responsible for answering "where to grasp," while force feedback is responsible for answering "how much force to use"; together with motion planning, they form a complete grasping pipeline—this is the specific meaning of the "coupling of perception and control" emphasized in the article.

The Community Value of Open-Source Humanoid Robots

From a community perspective, Asimov 1's progress carries significance beyond the single project. Open-source hardware allows researchers and students to obtain a real platform at low cost and enables different teams to compare algorithms on the same benchmark. As a frequently used end module, once an open-source design for the gripper takes shape, it can easily be borrowed or further modified by other projects.

The author's approach to sharing is also worth noting: continuously publishing progress in the form of a "Day 363" log both documents the development process and provides followers with a trackable roadmap. This kind of transparent iteration is precisely the core character that distinguishes the open-source robotics community from commercial product launches.

What to Watch for Next

Gripper testing is only the beginning. Directions worth observing going forward include: the grasping success rate and the range of manipulable objects, the material cost of the gripper solution, and whether standalone gripper design files will be released for the community to replicate. If Asimov 1 can achieve stable grasping and manipulation on an open-source basis, it will provide a truly usable entry point for individual developers to participate in humanoid robotics research.

One-line Summary

Open-source humanoid robot Asimov 1 begins testing new grippers on Day 363, moving from lifting toward grasping and manipulation.

Paragraph Summary

On Day 363, the open-source humanoid robot project Asimov 1 enters a new phase: the team is testing grippers with the goal of advancing the robot from merely pushing or lifting objects to truly grasping and interacting with them. The article notes that Asimov 1's defining features are being fully open source and built with off-the-shelf parts, with code hosted on GitHub, allowing ordinary people to replicate it by following the drawings and bill of materials—a sharp contrast to commercial humanoid robots that rely on custom actuators and closed supply chains. The author further analyzes that grasping is the second and harder threshold for humanoid robots after walking and balance, involving three challenges: end-effector complexity, the coupling of perception and control, and cost and maintainability. Gripper testing shows that the project's basic locomotion and arm structure have stabilized, and its focus is shifting toward practical usability.