Vehicle 1
Getting Around
A machine with one sensor and one motor should not feel alive. Braitenberg's trick is that it does not take much motion before an observer starts inventing motives anyway.
9 min read
The first Braitenberg vehicle is almost offensively small. It does not store a map. It does not classify the world. It does not decide between alternatives. It only couples one sensor to one motor and exposes that loop to a field.
That minimality is the point. Once the sensor reading modulates speed, and once the world is even slightly uneven, the path begins to look like temperament. It rushes, hesitates, overshoots, lingers, and rebounds. The mechanism is trivial. The interpretation is not.
Key idea
Vehicle 1 is not a study of intelligence. It is a study of how fast observers project intelligence onto movement. One scalar reading and one actuator already generate enough behavior for a story to begin.
One loop, no steering
In this first figure, warmer regions simply make the motor drive harder. The heading only changes because of drift and wall contact. That means any appearance of searching or curiosity is already being manufactured by the interaction between body and environment, not by a hidden decision process.
The path is the point. Warmer regions only change speed; drift and wall contact do the rest.
Temperament is often physics in disguise
The fastest way to make the same system look like a different personality is not to give it memory. It is to change the physical assumptions around it. Slightly more friction makes motion look reluctant. Slightly more noise makes it look nervous. Increase field strength and the same vehicle starts to feel driven.
This is a useful correction for modern software too. Behavior that gets described as preference, intent, or even policy is often the side effect of response curves, latency, damping, and environmental structure.
Friction and noise change the reading long before the mechanism changes.
Modern translation
Vehicle 1 is a reactive agent before the term existed. It is policy without representation. The agent does not carry a world model around. The world is read directly at the sensor and fed back into action.
The environment co-authors the behavior
If you only watch the body, it is easy to over-credit the vehicle. If you move the field around, the illusion loosens. The route is partly a property of the mechanism and partly a property of what the world is offering it to read.
In the editor below, move the warm source, add cold pockets, or erase one. The same loop behaves differently because the world is different, not because the vehicle became smarter.
Click the plate to move the warm source, add a cold source, or erase one.
Open sandbox
The full loop is exposed here: source layout, field strength, sensor gain, motor gain, trail, and run reset. This is not a big simulator because it does not need to be. Braitenberg's point survives contact with the browser precisely because the mechanism stays small.
What matters is that the threshold for apparent psychology is lower than intuition expects. The machine is not seeking, fearing, or preferring in any rich sense. We are reading those things into the motion because we are built to do that.
Open the full loop: source layout, field strength, gain, trail, and reset.