PerceptionLabPortable / app /nodes /antenna_evolution.py
Aluode's picture
Upload folder using huggingface_hub
3bb804c verified
"""
Antenna Evolution: Shape as Electromagnetic Interface
======================================================
The insight chain:
1. DNA is a fractal antenna (Blank & Goodman 2011) - structure determines bandwidth
2. Dendrites are frequency-tuned antennas (MIT, ephaptic coupling research)
3. Evolved shapes ARE antenna patterns - not decoration, but functional geometry
4. Multiple antennas with compatible shapes can COUPLE through field effects
This module implements:
- Organisms as antenna patterns with computable radiation characteristics
- Field coupling between organisms based on shape resonance
- An ecosystem where shapes "talk" to each other through their geometry
- Evolution that optimizes for both survival AND communication
The latent vector is the "DNA" that generates the antenna pattern.
The shape IS the antenna. The antenna determines what you can hear.
"""
import numpy as np
import cv2
from collections import deque
from scipy.fft import fft, ifft
from scipy.spatial.distance import cdist
# --- STRICT COMPATIBILITY IMPORTS ---
import __main__
try:
BaseNode = __main__.BaseNode
QtGui = __main__.QtGui
except AttributeError:
from PyQt6 import QtGui
class BaseNode:
def get_blended_input(self, name, mode): return None
def dna_to_antenna_pattern(dna, n_points=64):
"""
Convert DNA to antenna radiation pattern.
Uses Fourier synthesis - DNA encodes frequency components.
Returns: (boundary_points, frequency_response)
- boundary_points: the physical shape
- frequency_response: what frequencies this antenna can receive/transmit
"""
if dna is None or len(dna) < 8:
dna = np.zeros(32)
angles = np.linspace(0, 2*np.pi, n_points, endpoint=False)
# DNA as Fourier coefficients for the shape
n_harmonics = min(12, len(dna) // 2)
# Build the shape
radii = np.ones(n_points) * 50 # base radius
for k in range(n_harmonics):
amp = dna[k*2] * 15
phase = dna[k*2 + 1] * np.pi
harmonic = k + 1
radii += amp * np.cos(harmonic * angles + phase)
radii = np.clip(radii, 10, 100)
# The frequency response IS the DNA (Fourier coefficients)
# Normalized to unit energy
freq_response = np.abs(dna[:n_harmonics*2:2]) # Just the amplitudes
if np.sum(freq_response) > 0:
freq_response = freq_response / np.sum(freq_response)
# Convert to cartesian
cx, cy = 64, 64
points = np.array([(cx + r * np.cos(a), cy + r * np.sin(a))
for a, r in zip(angles, radii)])
return points, freq_response
def compute_antenna_coupling(dna1, dna2):
"""
Compute coupling strength between two antenna patterns.
Based on:
1. Frequency overlap (do they resonate at same frequencies?)
2. Impedance matching (complementary vs similar shapes)
Returns: coupling coefficient (0 to 1)
"""
_, freq1 = dna_to_antenna_pattern(dna1)
_, freq2 = dna_to_antenna_pattern(dna2)
# Pad to same length
max_len = max(len(freq1), len(freq2))
freq1 = np.resize(freq1, max_len)
freq2 = np.resize(freq2, max_len)
# Frequency overlap - dot product of normalized spectra
overlap = np.dot(freq1, freq2)
# Phase coherence - how aligned are their Fourier phases?
if len(dna1) >= 16 and len(dna2) >= 16:
phases1 = dna1[1:16:2] # odd indices = phases
phases2 = dna2[1:16:2]
phase_coherence = np.abs(np.mean(np.exp(1j * (phases1 - phases2))))
else:
phase_coherence = 0.5
# Combined coupling
coupling = overlap * 0.6 + phase_coherence * 0.4
return float(np.clip(coupling, 0, 1))
def compute_field_at_point(source_dna, source_pos, target_pos, time=0):
"""
Compute the electromagnetic field contribution from one antenna at a point.
The field strength depends on:
1. Distance (inverse square falloff)
2. Direction (antenna pattern is directional)
3. Time (oscillating field)
"""
dx = target_pos[0] - source_pos[0]
dy = target_pos[1] - source_pos[1]
distance = np.sqrt(dx*dx + dy*dy) + 1e-6
angle = np.arctan2(dy, dx)
# Get antenna pattern (angular gain)
points, freq_response = dna_to_antenna_pattern(source_dna, n_points=32)
# Angular index into pattern
pattern_idx = int((angle / (2*np.pi) + 0.5) * 32) % 32
# Radial extent at this angle = gain in this direction
center = np.array([64, 64])
gain = np.linalg.norm(points[pattern_idx] - center) / 50.0 # normalized
# Field strength with distance falloff
field_strength = gain / (1 + distance * 0.02)
# Oscillating component (superposition of frequencies)
oscillation = 0
for k, amp in enumerate(freq_response):
freq = (k + 1) * 0.5 # frequency in arbitrary units
oscillation += amp * np.sin(2 * np.pi * freq * time)
return field_strength * (1 + 0.3 * oscillation)
# =============================================================================
# Antenna Field Node - Visualizes the electromagnetic field of an organism
# =============================================================================
class AntennaFieldNode(BaseNode):
"""
Visualizes the electromagnetic field pattern of an organism.
The shape determines the radiation pattern - like viewing an antenna
in a near-field measurement chamber.
"""
NODE_CATEGORY = "Artificial Life"
NODE_COLOR = QtGui.QColor(100, 150, 255)
def __init__(self):
super().__init__()
self.node_title = "Antenna Field"
self.inputs = {
'dna': 'spectrum',
'frequency': 'signal' # Which frequency to visualize
}
self.outputs = {
'field_view': 'image',
'bandwidth': 'signal', # How many frequencies can it receive
'directivity': 'signal' # How focused is the pattern
}
self.time = 0.0
self.display = np.zeros((128, 128, 3), dtype=np.uint8)
def step(self):
dna = self.get_blended_input('dna', 'mean')
freq_select = self.get_blended_input('frequency', 'mean')
if dna is None:
dna = np.random.randn(32) * 0.3
if freq_select is None:
freq_select = 0.5
self.time += 0.1
# Get antenna pattern
points, freq_response = dna_to_antenna_pattern(dna)
# Compute field on a grid
self.display.fill(0)
field = np.zeros((128, 128))
for y in range(0, 128, 2):
for x in range(0, 128, 2):
f = compute_field_at_point(dna, (64, 64), (x, y), self.time)
field[y:y+2, x:x+2] = f
# Normalize and colorize
field = np.clip(field, 0, 2)
field_norm = (field / 2 * 255).astype(np.uint8)
# Color map: blue (weak) -> cyan -> green -> yellow (strong)
self.display[:,:,0] = np.clip(field_norm * 0.3, 0, 255).astype(np.uint8)
self.display[:,:,1] = np.clip(field_norm * 0.8, 0, 255).astype(np.uint8)
self.display[:,:,2] = np.clip(255 - field_norm, 0, 255).astype(np.uint8)
# Draw antenna outline
pts = points.astype(np.int32).reshape((-1, 1, 2))
cv2.polylines(self.display, [pts], True, (255, 255, 255), 1)
# Metrics
self.bandwidth = float(np.sum(freq_response > 0.05)) # Active bands
self.directivity = float(np.std(freq_response) * 10) # Pattern variation
def get_output(self, name):
if name == 'field_view': return self.display
if name == 'bandwidth': return self.bandwidth
if name == 'directivity': return self.directivity
return None
# =============================================================================
# Field Ecosystem Node - Multiple organisms coupling through field effects
# =============================================================================
class FieldEcosystemNode(BaseNode):
"""
An ecosystem where organisms influence each other through field coupling.
Each organism:
- Has a position in 2D space
- Radiates a field determined by its shape (DNA)
- Receives energy from other organisms' fields
- Evolves based on total energy received (fitness)
This creates selection pressure for:
- Shapes that can receive energy (good antennas)
- Shapes that couple well with neighbors (resonance)
- Possibly: shapes that can "communicate" information
"""
NODE_CATEGORY = "Artificial Life"
NODE_COLOR = QtGui.QColor(200, 100, 255)
def __init__(self):
super().__init__()
self.node_title = "Field Ecosystem"
self.inputs = {
'external_signal': 'spectrum', # Environmental broadcast (e.g., EEG)
'mutation_rate': 'signal'
}
self.outputs = {
'best_receiver_dna': 'spectrum',
'best_transmitter_dna': 'spectrum',
'ecosystem_view': 'image',
'avg_coupling': 'signal',
'generation': 'signal'
}
# Population
self.pop_size = 16
self.dna_len = 32
# Each organism: (dna, position, energy_received, energy_transmitted)
self.organisms = []
for _ in range(self.pop_size):
dna = np.random.randn(self.dna_len) * 0.5
pos = np.random.rand(2) * 100 + 14 # positions in 128x128 space
self.organisms.append({
'dna': dna,
'pos': pos,
'received': 0.0,
'transmitted': 0.0
})
self.gen = 0
self.time = 0.0
self.display = np.zeros((256, 256, 3), dtype=np.uint8)
self.coupling_history = deque(maxlen=100)
def step(self):
external = self.get_blended_input('external_signal', 'mean')
mutation_rate = self.get_blended_input('mutation_rate', 'mean')
if mutation_rate is None:
mutation_rate = 0.1
self.time += 0.1
# Reset energy accumulators
for org in self.organisms:
org['received'] = 0.0
org['transmitted'] = 0.0
# Compute pairwise field interactions
total_coupling = 0
n_pairs = 0
for i, org_i in enumerate(self.organisms):
for j, org_j in enumerate(self.organisms):
if i >= j:
continue
# Distance
dist = np.linalg.norm(org_i['pos'] - org_j['pos'])
# Antenna coupling (shape-based)
coupling = compute_antenna_coupling(org_i['dna'], org_j['dna'])
# Field strength falls off with distance
field_factor = 1.0 / (1 + dist * 0.05)
# Energy exchange
energy = coupling * field_factor
org_i['received'] += energy
org_j['received'] += energy
org_i['transmitted'] += energy
org_j['transmitted'] += energy
total_coupling += coupling
n_pairs += 1
# Add external signal reception
if external is not None and len(external) >= self.dna_len:
for org in self.organisms:
ext_coupling = compute_antenna_coupling(org['dna'], external[:self.dna_len])
org['received'] += ext_coupling * 2 # External signal is strong
# Record average coupling
if n_pairs > 0:
self.coupling_history.append(total_coupling / n_pairs)
# Evolution every N steps
if self.time % 3.0 < 0.15:
self._evolve_population(mutation_rate)
self.gen += 1
# Visualization
self._draw_ecosystem()
def _evolve_population(self, mutation_rate):
"""Selection and breeding based on energy received"""
# Sort by fitness (received energy)
sorted_orgs = sorted(self.organisms,
key=lambda o: o['received'],
reverse=True)
new_orgs = []
elite = max(2, int(self.pop_size * 0.25))
# Keep elite
for i in range(elite):
new_orgs.append({
'dna': sorted_orgs[i]['dna'].copy(),
'pos': sorted_orgs[i]['pos'].copy(),
'received': 0.0,
'transmitted': 0.0
})
# Breed rest
while len(new_orgs) < self.pop_size:
# Select parents from top half
p1 = sorted_orgs[np.random.randint(0, elite * 2)]
p2 = sorted_orgs[np.random.randint(0, elite * 2)]
# Crossover
alpha = np.random.rand(self.dna_len)
child_dna = p1['dna'] * alpha + p2['dna'] * (1 - alpha)
# Mutation
if np.random.rand() < 0.5:
child_dna += np.random.randn(self.dna_len) * mutation_rate
# Position: near parents with some spread
child_pos = (p1['pos'] + p2['pos']) / 2 + np.random.randn(2) * 10
child_pos = np.clip(child_pos, 14, 114)
new_orgs.append({
'dna': child_dna,
'pos': child_pos,
'received': 0.0,
'transmitted': 0.0
})
self.organisms = new_orgs
def _draw_ecosystem(self):
self.display.fill(10)
# Draw field lines between coupled organisms
for i, org_i in enumerate(self.organisms):
for j, org_j in enumerate(self.organisms):
if i >= j:
continue
coupling = compute_antenna_coupling(org_i['dna'], org_j['dna'])
if coupling > 0.3: # Only show strong couplings
p1 = (int(org_i['pos'][0] * 2), int(org_i['pos'][1] * 2))
p2 = (int(org_j['pos'][0] * 2), int(org_j['pos'][1] * 2))
intensity = int(coupling * 200)
cv2.line(self.display, p1, p2, (intensity//2, intensity, intensity//2), 1)
# Draw organisms
max_received = max(o['received'] for o in self.organisms) + 1e-6
for org in self.organisms:
# Position (scaled to 256x256)
cx = int(org['pos'][0] * 2)
cy = int(org['pos'][1] * 2)
# Get shape points
points, _ = dna_to_antenna_pattern(org['dna'], n_points=16)
# Scale and translate
points = (points - 64) * 0.3 + [cx, cy]
pts = points.astype(np.int32).reshape((-1, 1, 2))
# Color by received energy
energy_ratio = org['received'] / max_received
color = (50, int(100 + 155 * energy_ratio), int(100 + 100 * energy_ratio))
cv2.polylines(self.display, [pts], True, color, 1)
cv2.circle(self.display, (cx, cy), 2, color, -1)
# Labels
cv2.putText(self.display, f"Gen: {self.gen}", (5, 20),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (200, 200, 200), 1)
if len(self.coupling_history) > 0:
avg = np.mean(list(self.coupling_history)[-20:])
cv2.putText(self.display, f"Coupling: {avg:.2f}", (5, 40),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (200, 200, 200), 1)
def get_output(self, name):
if name == 'ecosystem_view': return self.display
if name == 'best_receiver_dna':
best = max(self.organisms, key=lambda o: o['received'])
return best['dna'].copy()
if name == 'best_transmitter_dna':
best = max(self.organisms, key=lambda o: o['transmitted'])
return best['dna'].copy()
if name == 'avg_coupling':
if len(self.coupling_history) == 0:
return 0.0
return float(np.mean(list(self.coupling_history)[-20:]))
if name == 'generation':
return float(self.gen)
return None
# =============================================================================
# Resonance Network Node - Organisms form a wireless neural network
# =============================================================================
class ResonanceNetworkNode(BaseNode):
"""
Organisms as nodes in a wireless neural network.
Information propagates through field coupling.
An input signal at one organism propagates to others
based on their antenna coupling coefficients.
This is ephaptic coupling, scaled up to the ecosystem level.
"""
NODE_CATEGORY = "Artificial Life"
NODE_COLOR = QtGui.QColor(255, 150, 100)
def __init__(self):
super().__init__()
self.node_title = "Resonance Network"
self.inputs = {
'input_signal': 'signal', # Signal injected into network
'organism_dnas': 'spectrum', # DNA patterns of network nodes
'topology': 'signal' # 0=ring, 1=random, 2=fully connected
}
self.outputs = {
'output_signal': 'signal', # Signal at output node
'propagation_view': 'image',
'network_coherence': 'signal' # How synchronized is the network
}
self.n_nodes = 8
self.node_states = np.zeros(self.n_nodes)
self.node_dnas = [np.random.randn(32) * 0.5 for _ in range(self.n_nodes)]
# Precompute coupling matrix
self.coupling_matrix = np.zeros((self.n_nodes, self.n_nodes))
self._update_coupling_matrix()
self.display = np.zeros((128, 128, 3), dtype=np.uint8)
self.history = deque(maxlen=50)
def _update_coupling_matrix(self):
for i in range(self.n_nodes):
for j in range(self.n_nodes):
if i != j:
self.coupling_matrix[i, j] = compute_antenna_coupling(
self.node_dnas[i], self.node_dnas[j]
)
def step(self):
input_sig = self.get_blended_input('input_signal', 'mean')
new_dnas = self.get_blended_input('organism_dnas', 'mean')
# Update DNAs if provided
if new_dnas is not None and len(new_dnas) >= 32:
# Use chunks of the input as different node DNAs
for i in range(min(self.n_nodes, len(new_dnas) // 32)):
self.node_dnas[i] = new_dnas[i*32:(i+1)*32]
self._update_coupling_matrix()
if input_sig is None:
input_sig = np.sin(len(self.history) * 0.2) # Default oscillation
# Inject signal at first node
self.node_states[0] = input_sig
# Propagate through network (one step of diffusion)
new_states = np.zeros(self.n_nodes)
for i in range(self.n_nodes):
# Self-decay
new_states[i] = self.node_states[i] * 0.8
# Input from coupled neighbors
for j in range(self.n_nodes):
if i != j:
new_states[i] += self.node_states[j] * self.coupling_matrix[j, i] * 0.3
self.node_states = np.tanh(new_states) # Nonlinearity
self.history.append(self.node_states.copy())
# Visualization
self._draw_network()
def _draw_network(self):
self.display.fill(10)
# Node positions in a circle
cx, cy = 64, 64
radius = 45
positions = []
for i in range(self.n_nodes):
angle = i * 2 * np.pi / self.n_nodes - np.pi/2
x = int(cx + radius * np.cos(angle))
y = int(cy + radius * np.sin(angle))
positions.append((x, y))
# Draw coupling lines
for i in range(self.n_nodes):
for j in range(i+1, self.n_nodes):
coupling = self.coupling_matrix[i, j]
if coupling > 0.2:
intensity = int(coupling * 150)
cv2.line(self.display, positions[i], positions[j],
(intensity//2, intensity, intensity//2), 1)
# Draw nodes
for i, (x, y) in enumerate(positions):
# Size by state amplitude
size = int(5 + abs(self.node_states[i]) * 10)
# Color by state sign
if self.node_states[i] > 0:
color = (50, 200, 50) # Green = positive
else:
color = (50, 50, 200) # Blue = negative
cv2.circle(self.display, (x, y), size, color, -1)
cv2.circle(self.display, (x, y), size, (200, 200, 200), 1)
# Input/output markers
cv2.putText(self.display, "IN", (positions[0][0]-8, positions[0][1]-12),
cv2.FONT_HERSHEY_SIMPLEX, 0.3, (255, 255, 255), 1)
cv2.putText(self.display, "OUT", (positions[self.n_nodes//2][0]-10,
positions[self.n_nodes//2][1]-12),
cv2.FONT_HERSHEY_SIMPLEX, 0.3, (255, 255, 255), 1)
# Coherence indicator
coherence = self._compute_coherence()
bar_width = int(coherence * 100)
cv2.rectangle(self.display, (14, 118), (14 + bar_width, 124),
(100, 200, 100), -1)
def _compute_coherence(self):
"""How synchronized are the node states?"""
if len(self.history) < 10:
return 0.0
recent = np.array(list(self.history)[-10:])
# Coherence = how correlated are the oscillations
correlations = []
for i in range(self.n_nodes):
for j in range(i+1, self.n_nodes):
corr = np.corrcoef(recent[:, i], recent[:, j])[0, 1]
if not np.isnan(corr):
correlations.append(abs(corr))
if len(correlations) == 0:
return 0.0
return float(np.mean(correlations))
def get_output(self, name):
if name == 'output_signal':
return float(self.node_states[self.n_nodes // 2])
if name == 'propagation_view':
return self.display
if name == 'network_coherence':
return self._compute_coherence()
return None
# =============================================================================
# Fractal Antenna Node - DNA-like self-similar structure
# =============================================================================
class FractalAntennaNode(BaseNode):
"""
Generates fractal antenna patterns inspired by the DNA structure.
From Blank & Goodman (2011):
- DNA has multiple scales of coiling (1nm helix → 10nm fiber → 30nm solenoid → 200nm tube)
- Each scale resonates with different frequencies
- Self-similarity creates broadband reception
This node generates shapes with explicit fractal structure.
"""
NODE_CATEGORY = "Artificial Life"
NODE_COLOR = QtGui.QColor(255, 200, 50)
def __init__(self):
super().__init__()
self.node_title = "Fractal Antenna"
self.inputs = {
'seed_dna': 'spectrum',
'fractal_depth': 'signal', # 1-4 levels of self-similarity
'base_frequency': 'signal'
}
self.outputs = {
'fractal_dna': 'spectrum',
'antenna_view': 'image',
'bandwidth': 'signal' # Number of frequency bands
}
self.dna = np.zeros(64)
self.display = np.zeros((128, 128, 3), dtype=np.uint8)
def step(self):
seed = self.get_blended_input('seed_dna', 'mean')
depth = self.get_blended_input('fractal_depth', 'mean')
base_freq = self.get_blended_input('base_frequency', 'mean')
if seed is None:
seed = np.random.randn(16) * 0.5
if depth is None:
depth = 3
if base_freq is None:
base_freq = 1.0
depth = int(np.clip(depth, 1, 4))
# Generate fractal DNA
# Each level adds scaled copies of the base pattern
self.dna = np.zeros(64)
base = np.resize(seed, 16)
for level in range(depth):
scale = 2 ** level
freq_mult = base_freq * scale
# Add base pattern at this scale
for i, val in enumerate(base):
idx = int(i * scale) % 64
self.dna[idx] += val / scale # Amplitude decreases with scale
# Normalize
if np.max(np.abs(self.dna)) > 0:
self.dna = self.dna / np.max(np.abs(self.dna))
# Visualization
self._draw_fractal_antenna(depth)
def _draw_fractal_antenna(self, depth):
self.display.fill(10)
cx, cy = 64, 64
# Draw antenna at multiple scales
colors = [(100, 200, 255), (100, 255, 200), (255, 200, 100), (255, 100, 200)]
for level in range(depth):
scale = 2 ** level
radius = 20 + level * 15
n_points = 8 * (level + 1)
points = []
for i in range(n_points):
angle = i * 2 * np.pi / n_points
# Modulate radius by DNA
dna_idx = int(i * 64 / n_points) % 64
r_mod = radius + self.dna[dna_idx] * 10 * (depth - level)
x = int(cx + r_mod * np.cos(angle))
y = int(cy + r_mod * np.sin(angle))
points.append((x, y))
# Draw this level
color = colors[level % len(colors)]
pts = np.array(points, dtype=np.int32).reshape((-1, 1, 2))
cv2.polylines(self.display, [pts], True, color, 1)
# Labels
cv2.putText(self.display, f"Depth: {depth}", (5, 15),
cv2.FONT_HERSHEY_SIMPLEX, 0.4, (200, 200, 200), 1)
cv2.putText(self.display, f"Bands: {depth * 4}", (5, 30),
cv2.FONT_HERSHEY_SIMPLEX, 0.4, (200, 200, 200), 1)
def get_output(self, name):
if name == 'fractal_dna':
return self.dna.copy()
if name == 'antenna_view':
return self.display
if name == 'bandwidth':
# Count significant frequency components
fft_result = np.abs(fft(self.dna))
return float(np.sum(fft_result > 0.1))
return None