The Problems With Humanoid Robots

🌈 Abstract

The article discusses the author's views on the development of humanoid robots, highlighting the challenges and limitations of this technology. The author argues that while the engineering behind humanoid robots is impressive, they are not the most practical solution for most real-world tasks.

🙋 Q&A

[01] Problems with Humanoid Robots

1. What are the three specific problems with humanoid robots that the author discusses?

• The first problem is the difficulty in developing robust controllers for humanoid robots, which the author believes will be overcome with continued advances in AI.
• The second problem is the high cost of complex hardware engineering combined with the unsolved AI problems, which may require significant investor funding.
• The third problem is that the author believes humanoids are the wrong solution for most tasks, as wheels and other form factors are more practical for most commercial and industrial applications.

2. Why does the author think the controls problem for humanoid robots is as hard or even harder than developing self-driving cars? The author explains that self-driving cars are passively stable and don't have to deal with issues like mass sloshing that can significantly complicate the controls problem for a humanoid robot carrying an object. Humans use a complex system of muscles, including neck muscles, to stabilize our bodies, which is very difficult to replicate in a robot.

3. What are the advantages of using wheels and other form factors over humanoid robots for most tasks? The author argues that wheels are the right answer for logistics, manufacturing, hospitals, airports, and most commercial environments. Wheels provide passive stability with at least 3 or 4 points of contact on the ground, and keep the payload inside the cone of stability rather than cantilevered in front of the robot.

[02] The Author's Approach at Cobot

1. What is the approach the author's company, Cobot, is taking with their collaborative robot? The author states that Cobot is not trying to solve open-ended AI problems, but is instead taking a pragmatic approach of using wheels, 4 points of ground contact, and keeping payloads within the cone of stability. The goal is to bring the benefits of robots outside of controlled environments and into the world around us in a trustworthy and collaborative way.

2. How does the author's approach at Cobot differ from the development of humanoid robots? The author contrasts Cobot's pragmatic approach of leveraging existing technology like wheels, rather than trying to solve the complex challenges of humanoid robotics. The focus is on deploying current state-of-the-art robots at scale quickly, rather than pursuing open-ended AI problems.