# The Control Layer — A Guide

**v4, August 13, 2026**
*From the work of Art Code Outdoors*
*Written by Bryan Carter and Claude Opus 4.6. Errors are Bryan's own.*

**Purpose:** This document explains the control layer of the Andromeda architecture for a general audience. It is a companion to the Layman's Guide (which covers the full architecture) and focuses specifically on the reactive substrate that the learning, attention, and MIRROR layers build upon. Readers do not need an engineering background. A reader who finishes this document will know what the control layer is, why it works the way it does, and where its boundary is — but would not be able to build one. For that, see the companion *Control Layer Technical Specification*.

**License:** MIT-0

> Every architecture in the Walter → Braitenberg → Brooks → Tilden lineage is a control layer architecture. Andromeda begins where these architectures stop — at the boundary where reactive behavior cannot adapt within a single lifetime. This document describes the foundation, not the building. Building from this document alone produces a Braitenberg vehicle or a BEAM robot, both real and valuable, but neither is Andromeda. The full architecture requires the learning layer, attention layer, MIRROR mechanism, and Universal Constructor described in the Framework document.

---

## One Sensor, One Motor, One Wire

A sensor reads a number: 0 (nothing) to 100 (maximum). A motor takes a number: 0 (stopped) to 100 (full speed). Between them is a wire with a rule.

If the wire is excitatory, the rule is: motor = sensor. The brighter the light, the faster the motor spins.

If the wire is inhibitory, the rule is: motor = 100 minus sensor. The brighter the light, the slower the motor spins.

That is the entire inner life of the connection. The sensor does not know what made the light. It reports a number. The motor does not know why it is spinning. It receives a number. Nobody inside the system knows what is happening from the outside. Heinz von Foerster called this the undifferentiated encoding: "The response of a nerve cell does not encode the physical nature of the agents that caused its response. Encoded is only 'how much' at this point on my body, but not 'what.'"

## Two Sensors, Two Motors, Four Personalities

Valentino Braitenberg imagined the simplest possible robot: two light sensors, two motors, and wires between them. Two design choices — where the wires go and what kind they are — produce four distinct "personalities."

**Wiring** can be uncrossed (left sensor → left motor, right sensor → right motor) or crossed (left sensor → right motor, right sensor → left motor).

**Connection type** can be excitatory (more signal = more speed) or inhibitory (more signal = less speed).

Four combinations:

**Fear** — uncrossed, excitatory. Each motor's rule: motor = sensor (same side). When light hits the left sensor, the left motor speeds up. The robot swerves away. The brighter the light, the faster it runs. Rule per motor: "my speed equals my sensor's reading."

**Aggression** — crossed, excitatory. Each motor's rule: motor = opposite sensor. When light hits the left sensor, the *right* motor speeds up. The robot turns toward the light and accelerates. It charges the source at full speed. Rule per motor: "my speed equals the opposite sensor's reading."

**Explorer** — uncrossed, inhibitory. Each motor's rule: motor = 100 minus sensor (same side). Far from light, sensors read low, both motors run fast — the robot zooms around. Near light, sensors read high, motors slow down — the robot crawls past without stopping, then speeds up again when it is past. Rule per motor: "my speed equals 100 minus my sensor's reading."

**Love** — crossed, inhibitory. Each motor's rule: motor = 100 minus opposite sensor. The robot turns toward the light (because the side facing the light slows down more), and decelerates as it approaches. It comes to rest near the source. Rule per motor: "my speed equals 100 minus the opposite sensor's reading."

Four behaviors. No code, no decisions, no if-then statements, no state machines. Just the physics of two numbers going through two wires on a body with two sides. Braitenberg named them after emotions because the behaviors *look* like emotions from the outside. From the inside, a sensor reads 73 and a motor computes 100 minus 73 equals 27. That is the complete experience.

Braitenberg's greater point: biological wiring is not really any different when you look at it under the microscope. There are just a lot more rules.

## Boids — The Same Principle in a Flock

Craig Reynolds (1986) showed the same idea working in groups. Three rules, each individual follows using only local information:

1. **Separation** — steer away from neighbors that are too close.
2. **Alignment** — steer toward the average heading of nearby neighbors.
3. **Cohesion** — steer toward the average position of nearby neighbors.

No leader. No bird knows what the flock looks like. Each one reacts only to what is nearby. Flocking — the coordinated, flowing, organic movement of hundreds of individuals — emerges from three local rules.

Every Braitenberg vehicle rule could be restated as a boids rule. Fear: "move away from bright things, faster when they're close." Love: "move toward bright things, slower when they're close." The format is the same: a local reactive rule that produces complex-looking behavior as a side effect of running on a physical body in a physical world.

A single Braitenberg vehicle has one sensor pair and one behavior. The jump from single-vehicle to flocking is not adding complexity to any rule. It is running multiple simple rules on the same body at the same time. Three sensor pairs — separation, cohesion, alignment — all wired to the same two thrusters. Each pair pushes its modulation into the thruster independently. The thruster does not know there are three wires. It has one charge level — the sum of everything flowing in and out. One number. It fires based on that one number.

No rule knows about any other rule. No rule decides which one wins. The physics sums the modulations, the body moves, the environment changes, the sensors read new numbers, and the cycle continues. Flocking is what that looks like from the outside.

> "Each rule is like a primitive reflex. Many reflexes together make a behavior."

## The Nervous Net — BEAM Robotics

Mark Tilden built real robots on this principle. BEAM robotics (Biology, Electronics, Aesthetics, Mechanics) does not program robots. It builds a nervous system from analog electronics, and behavior emerges from the physics of the circuit.

The building block is the Nv — the nervous net. Two or more cells wired in mutual inhibition. The simplest is a bicore: two cells, each inhibiting the other. Cell A charges, hits threshold, fires, suppresses Cell B. Cell A exhausts, recovers. Cell B, no longer suppressed, fires, suppresses Cell A. Back and forth. This is the heartbeat. It runs on capacitor charge/discharge cycles — no clock, no software, no instructions. Pure analog oscillation.

Connect a bicore to two legs: alternating leg movement. Walking. A three-cell ring produces a three-phase gait. The robot does not decide to walk. The circuit oscillates, the legs are connected to the circuit, and walking happens.

Sensors do not tell the robot what to do. They modulate the oscillation. A photoresistor on one side changes the charge rate of the cell it is connected to. More light means that cell charges faster, fires sooner, the oscillation becomes asymmetric, the robot turns. The robot does not "see" the light, process it, and decide to turn. The light physically changes the timing of an already-running oscillator. Steering is a side effect of asymmetric oscillation.

The heartbeat comes first. Everything else rides on top. Remove the heartbeat and the robot is dead — no amount of sensor input can make a stopped oscillator move. This is the fundamental principle: the nervous net oscillates, and sensors modulate the oscillation.

## W. Grey Walter's Descendants

W. Grey Walter built Elmer (1948) and Elsie (1949) — electromechanical tortoises with light sensors, contact shells, and simple reflex circuits. They exhibited phototaxis, obstacle avoidance, and what Walter called "recognition" of their own reflection. They were the first autonomous robots.

Every autonomous machine since descends from them:

- **Rodney Brooks' Subsumption Architecture**: Genghis (1989), Attila (1991), and their descendants — Mars Pathfinder's Sojourner (1997), iRobot's PackBot (2001), the original Roomba (2002). Layered reactive behaviors, no central planner, no world model.
- **Mark Tilden's BEAM Robotics**: Walkman (1991), Unibug 1.0 (LANL), SatBot (LANL Space), Solarbotics kits, WowWee's B.I.O. Bugs (2001) and Robosapien (2004). Analog nervous nets, no software at all.
- **Game AI**: the Xenomorph in Alien: Isolation (2014), the Flood in Halo. Behavior trees and reactive systems producing emergent behavior from authored rules.

All of these are the same thing expressed at different scales: local reactive rules, no prediction, no memory, no imagination. Fear, Aggression, Explorer, Love — with a lot more rules. From Braitenberg's two-sensor toy to a Roomba navigating a living room, the principle is identical. The behavior looks intelligent from the outside. From the inside, a sensor reads a number and a motor receives a number.

These are all control layer.

## What the Control Layer Cannot Do

A Braitenberg Fear vehicle flees from light. If you put it in a room with a candle, it runs to the dark corner. Every time. If you move the candle, it runs to the new dark corner. Every time. It never learns. It never remembers. It cannot predict that the candle will move. It cannot imagine what the room would look like with the candle somewhere else.

A moth flies toward flame because its phototaxis wiring says "turn toward the brightest thing." The moth cannot learn that flame kills moths. It cannot override the reflex. Evolution could breed a line of moths that avoid flame — but that takes generations. The individual moth that is flying toward the candle right now has no mechanism to stop.

This is the boundary of the control layer. It produces behavior that is adaptive in the environment the wiring was built for. It cannot adapt within a single lifetime. It cannot predict. It cannot imagine alternatives. It responds to what is happening right now, based on wiring that was set before it was born.

An enormous amount of capable, robust, adaptive-looking behavior lives within this boundary. A Roomba navigates furniture. A PackBot traverses rubble. A game AI hunts a player through a space station. All of it is reactive. All of it is "how much on my body" going through wires to motors. All of it is control layer.

## What Goes On Top

The Andromeda architecture starts with the control layer — a BEAM nervous net, the same lineage that descends from Walter's tortoises. The robot oscillates, sensors modulate the oscillation, reflexive behavior emerges from wiring. This is the foundation. It is well-established, well-tested, and it works.

What Andromeda adds is the machinery for the moth to learn about the candle.

The learning layer observes the control layer's activity — the full pattern of sensor readings and motor states — and learns sequences. Not single moments, but sequences of moments. It predicts what comes next. It does not control anything. It watches and predicts.

The attention layer sits between the learning layer and the control layer. It gates what gets in (which control layer activity the learning layer observes) and what gets out (which predictions reach the control layer). When a prediction is bad — the learning layer encounters something it has never seen and cannot predict — the attention layer detects the failure and squelches the prediction before it can affect the control layer.

The MIRROR mechanism is where the predictions re-enter the control layer as synthetic sensory input. The control layer cannot tell the difference between real sensor data and a prediction. It responds to both identically. This is how the learning layer acts on the future despite controlling nothing: it changes what the control layer perceives.

The moth with these layers flies toward the candle (control layer reflex), feels the heat (sensor modulation), learns the sequence "bright light → heat → pain" (learning layer), and the next time it sees a bright light, the learning layer predicts pain and the MIRROR mechanism feeds that prediction back to the control layer. The control layer reacts to the predicted pain as if it were real — and the moth turns away. Within its own lifetime. No breeding required.

That is what goes on top.

## Why Not Hand-Wire It?

A Braitenberg vehicle with four rules — separation, cohesion, alignment, wall avoidance — can be tuned by hand. You fiddle with the gains, watch the boids, adjust. With patience, you get something that works.

Now consider Alien: Isolation's Xenomorph. Over 100 behavior tree nodes. Each with sensors. Each interacting with every other through the same physical body and environment. Creative Assembly hand-tuned that over years of development, and even game developers at that level often resort to genetic algorithms to find the right balance.

The problem is the superposition. When many reflexes sum on the same motors, adjusting one coefficient changes the behavior produced by all the others. An RTS that adjusts deer reproduction rate finds crops failing and NPC villages starving. A shooter that introduces a new weapon breaks the entire sandbox. Each patch risks breaking something else. Even when balance is achieved, it holds only until contact with the player — someone finds the edge case, and the patching cycle restarts.

At the scale of a real organism — thousands of cells, millions of connections, hundreds of interacting reflexes — no human can hand-tune the coefficients. The interaction space is too large. The only way to find values that balance against each other is to let something search the space for you.

Von Neumann asked: what if the computer could program itself? He had no training data and no corpus of code. So he designed the universal constructor — a machine that builds new machines from a tape, with mutation and selection as the search mechanism. The trade-off is that evolution produces solutions that look strange from the outside — the platypus is working, but it is not what a human designer would choose. The payoff is that it searches a space no human can navigate.

That is why the Andromeda architecture evolves its wiring rather than specifying it. The control layer — Braitenberg vehicles, boids, the entire Elmer/Elsie descendant tree — is the part that works at small scale with hand-tuning. Andromeda's contribution is the machinery to evolve the wiring at scales where hand-tuning breaks down, and the learning layer on top that adapts within a single lifetime where evolution is too slow.

---

*The complete document bundle is available at kitchencloset.com/realstuff/andromeda/. The architecture is the work of Art Code Outdoors; this document is Bryan Carter's record of the designer's disclosures, and errors in it are Bryan's own. For contact information, see the Framework document.*