Our Approach

Back to First Principles

Robots has progressed significantly in recent times, but the dream of a general-purpose robot that has human capabilities and can safely operate in human environments hasn’t materialized So far, these “aren’t the droids your looking for”. We are looking at this problem differently.

Elon Musk has revolutionized transportation, power generation, and rocketry amongst others. To do this he went back to first principles. Why are rockets so expensive? It’s not the fuel, it’s the fact that they are single use. Make rockets reusable like planes. Why aren’t electric cars feasible? Because the batteries weigh too much you have to reduce the weight of the car. Apply the light construction techniques of rocketry to cars. Elon both went back to first principles and focused on cross-discipline solutions. At Forged Droids we are taking a similar approach.

Principle 1: In a world built for humans, we need droids with human capabilities.

Let’s face it, if we want droids to move out of their controlled, caged environments into the designed-for-humans world, droids need to be as capable as humans. Everything we build is designed for people: doorways, chairs, tools, cars, etc. To function in the designed-for-humans world droids need to be relatively the same in height, with similar hands, similar legs, similar strength, similar sensing, etc. These droids need to be able to work were people do safely and perform tasks previously only done by people.

Principle 2: Beat the human benchmark for structure, strength, and sensing.

How can we beat human structure, strength, joints, and sensing? First, understand what the human baseline is and then look across other domains for higher performing solutions. For instance, when we look at the weight relative to the strength of bone we find many solutions that can outperform it.

Principle 3: Lighter is safer, more power efficient, and less expensive.

SWAP-C stands for size, weight, power, and cost. These variables are interrelated. The heavier the droid the more materials go into making it. If the droid is heavier it takes larger actuators to run it. Larger actuators require bigger batteries. A 400lb droid falling on you is a much bigger threat than a 60lb droid. Generally lighter is safer and less expensive.

Principle 4: Beat the human benchmark for control.

A big part of droid development is getting the real-time control right. We find a magic number for low-level control is between 10ms and 20ms. This works whether controlling a helicopter or getting your hand off a hot stove. The most successful droids have well designed real-time behavior between force sensing and variable force actuation. We also find the higher-level cognition happens in the 100ms to 200ms range. This is the time it takes to identify objects around you and act.

Principle 5: Train one droid you train them all.

First, we want to create human training as augmented reality. This training will show the human what actions to take, in what order, and verify they were done correctly. Next, we will use this training for the robot. By generalizing the training (grasp, move, put in a screw etc.), if one droid learns it all droids learn it.