modules/swarm_intelligence.py

"""
Sistema de Inteligencia de Enjambre
Implementa algoritmos de optimización colectiva inspirados en comportamientos naturales
"""

import numpy as np
import random
from typing import Dict, List, Tuple, Callable, Any, Optional
from dataclasses import dataclass
from abc import ABC, abstractmethod
from concurrent.futures import ThreadPoolExecutor
import threading
import time
from datetime import datetime

@dataclass
class Agent:
    """Agente individual en el enjambre"""
    id: str
    position: np.ndarray
    velocity: np.ndarray
    best_position: np.ndarray
    best_fitness: float
    fitness: float
    role: str = "explorer"
    
@dataclass 
class SwarmState:
    """Estado actual del enjambre"""
    global_best_position: np.ndarray
    global_best_fitness: float
    agents: List[Agent]
    iteration: int
    convergence_rate: float
    diversity_measure: float

class OptimizationAlgorithm(ABC):
    """Clase base para algoritmos de optimización"""
    
    @abstractmethod
    def update_agent(self, agent: Agent, swarm_state: SwarmState) -> Agent:
        pass
    
    @abstractmethod
    def get_algorithm_name(self) -> str:
        pass

class ParticleSwarmOptimization(OptimizationAlgorithm):
    """Algoritmo de Optimización por Enjambre de Partículas"""
    
    def __init__(self, w: float = 0.729, c1: float = 1.494, c2: float = 1.494):
        self.w = w    # factor de inercia
        self.c1 = c1  # coeficiente cognitivo
        self.c2 = c2  # coeficiente social
    
    def update_agent(self, agent: Agent, swarm_state: SwarmState) -> Agent:
        """Actualiza posición y velocidad de un agente PSO"""
        r1, r2 = random.random(), random.random()
        
        # Actualizar velocidad
        cognitive_component = self.c1 * r1 * (agent.best_position - agent.position)
        social_component = self.c2 * r2 * (swarm_state.global_best_position - agent.position)
        
        agent.velocity = (self.w * agent.velocity + 
                         cognitive_component + 
                         social_component)
        
        # Actualizar posición
        agent.position = agent.position + agent.velocity
        
        return agent
    
    def get_algorithm_name(self) -> str:
        return "Particle Swarm Optimization"

class AntColonyOptimization(OptimizationAlgorithm):
    """Algoritmo de Optimización por Colonia de Hormigas"""
    
    def __init__(self, alpha: float = 1.0, beta: float = 2.0, rho: float = 0.1):
        self.alpha = alpha  # importancia del feromona
        self.beta = beta    # importancia de la heurística
        self.rho = rho      # tasa de evaporación
        self.pheromone_matrix = None
    
    def update_agent(self, agent: Agent, swarm_state: SwarmState) -> Agent:
        """Actualiza posición basada en feromonas"""
        if self.pheromone_matrix is None:
            self._initialize_pheromone_matrix(len(agent.position))
        
        # Selección probabilística basada en feromonas
        probabilities = self._calculate_probabilities(agent.position)
        new_position = self._select_next_position(agent.position, probabilities)
        
        agent.position = new_position
        return agent
    
    def _initialize_pheromone_matrix(self, size: int):
        """Inicializa matriz de feromonas"""
        self.pheromone_matrix = np.ones((size, size)) * 0.1
    
    def _calculate_probabilities(self, position: np.ndarray) -> np.ndarray:
        """Calcula probabilidades de transición"""
        if self.pheromone_matrix is None:
            return np.ones(len(position)) / len(position)
        
        # Simplificación: usa feromonas locales
        pheromones = np.diag(self.pheromone_matrix)
        heuristic = 1.0 / (1.0 + np.abs(position))  # heurística simple
        
        numerator = (pheromones ** self.alpha) * (heuristic ** self.beta)
        return numerator / np.sum(numerator)
    
    def _select_next_position(self, current_pos: np.ndarray, 
                            probabilities: np.ndarray) -> np.ndarray:
        """Selecciona siguiente posición"""
        # Movimiento basado en probabilidades
        direction = np.random.choice(len(probabilities), p=probabilities)
        step = np.zeros_like(current_pos)
        step[direction % len(current_pos)] = np.random.normal(0, 0.1)
        
        return current_pos + step
    
    def get_algorithm_name(self) -> str:
        return "Ant Colony Optimization"

class GeneticAlgorithm(OptimizationAlgorithm):
    """Algoritmo Genético"""
    
    def __init__(self, mutation_rate: float = 0.1, crossover_rate: float = 0.8):
        self.mutation_rate = mutation_rate
        self.crossover_rate = crossover_rate
    
    def update_agent(self, agent: Agent, swarm_state: SwarmState) -> Agent:
        """Actualiza agente mediante operaciones genéticas"""
        # Selección de padres
        parents = self._select_parents(swarm_state.agents)
        
        # Cruzamiento
        if random.random() < self.crossover_rate and len(parents) >= 2:
            agent.position = self._crossover(parents[0].position, parents[1].position)
        
        # Mutación
        if random.random() < self.mutation_rate:
            agent.position = self._mutate(agent.position)
        
        return agent
    
    def _select_parents(self, agents: List[Agent]) -> List[Agent]:
        """Selección de padres por torneo"""
        tournament_size = min(3, len(agents))
        tournament = random.sample(agents, tournament_size)
        return sorted(tournament, key=lambda a: a.fitness, reverse=True)[:2]
    
    def _crossover(self, parent1: np.ndarray, parent2: np.ndarray) -> np.ndarray:
        """Cruzamiento uniforme"""
        mask = np.random.random(len(parent1)) < 0.5
        child = np.where(mask, parent1, parent2)
        return child
    
    def _mutate(self, individual: np.ndarray) -> np.ndarray:
        """Mutación gaussiana"""
        mutation_mask = np.random.random(len(individual)) < self.mutation_rate
        mutations = np.random.normal(0, 0.1, len(individual))
        return individual + mutation_mask * mutations
    
    def get_algorithm_name(self) -> str:
        return "Genetic Algorithm"

class ConsensusBuilder:
    """Constructor de consenso para decisiones colectivas"""
    
    def __init__(self, convergence_threshold: float = 0.01):
        self.convergence_threshold = convergence_threshold
        self.voting_history = []
    
    def build_consensus(self, agent_opinions: List[np.ndarray]) -> Dict[str, Any]:
        """Construye consenso a partir de opiniones de agentes"""
        if not agent_opinions:
            return {"consensus": None, "confidence": 0.0}
        
        # Calcula centroide como consenso inicial
        consensus = np.mean(agent_opinions, axis=0)
        
        # Mide dispersión para evaluar confianza
        distances = [np.linalg.norm(opinion - consensus) 
                    for opinion in agent_opinions]
        avg_distance = np.mean(distances)
        confidence = 1.0 / (1.0 + avg_distance)
        
        # Identifica agentes disidentes
        outliers = [i for i, dist in enumerate(distances) 
                   if dist > 2 * np.std(distances)]
        
        # Proceso iterativo de refinamiento
        if len(outliers) > 0 and len(agent_opinions) > 3:
            # Recalcula consenso sin outliers extremos
            filtered_opinions = [op for i, op in enumerate(agent_opinions) 
                               if i not in outliers[:len(outliers)//2]]
            if filtered_opinions:
                refined_consensus = np.mean(filtered_opinions, axis=0)
                consensus = 0.7 * consensus + 0.3 * refined_consensus
        
        return {
            "consensus": consensus,
            "confidence": confidence,
            "participation_rate": len(agent_opinions),
            "outliers": len(outliers),
            "convergence_achieved": avg_distance < self.convergence_threshold
        }

class SwarmIntelligence:
    """Sistema principal de inteligencia de enjambre"""
    
    def __init__(self, 
                 swarm_size: int = 50,
                 dimensions: int = 10,
                 max_iterations: int = 100):
        self.swarm_size = swarm_size
        self.dimensions = dimensions
        self.max_iterations = max_iterations
        
        # Algoritmos disponibles
        self.algorithms = {
            'pso': ParticleSwarmOptimization(),
            'aco': AntColonyOptimization(),
            'ga': GeneticAlgorithm()
        }
        
        self.consensus_builder = ConsensusBuilder()
        self.swarm_state = None
        self.optimization_history = []
        
        # Control de ejecución
        self._running = False
        self._lock = threading.Lock()
    
    def initialize_swarm(self, 
                        bounds: Tuple[float, float] = (-10, 10),
                        fitness_function: Optional[Callable] = None) -> SwarmState:
        """Inicializa el enjambre"""
        agents = []
        
        for i in range(self.swarm_size):
            # Posición inicial aleatoria
            position = np.random.uniform(bounds[0], bounds[1], self.dimensions)
            velocity = np.random.uniform(-1, 1, self.dimensions)
            
            agent = Agent(
                id=f"agent_{i}",
                position=position,
                velocity=velocity,
                best_position=position.copy(),
                best_fitness=float('-inf'),
                fitness=float('-inf'),
                role=self._assign_role(i)
            )
            
            # Evaluar fitness inicial
            if fitness_function:
                agent.fitness = fitness_function(agent.position)
                agent.best_fitness = agent.fitness
            
            agents.append(agent)
        
        # Encontrar mejor global
        best_agent = max(agents, key=lambda a: a.fitness)
        
        self.swarm_state = SwarmState(
            global_best_position=best_agent.position.copy(),
            global_best_fitness=best_agent.fitness,
            agents=agents,
            iteration=0,
            convergence_rate=0.0,
            diversity_measure=self._calculate_diversity(agents)
        )
        
        return self.swarm_state
    
    def optimize(self, 
                fitness_function: Callable[[np.ndarray], float],
                algorithm: str = 'pso',
                bounds: Tuple[float, float] = (-10, 10),
                parallel: bool = True) -> Dict[str, Any]:
        """Ejecuta optimización con el algoritmo especificado"""
        
        if algorithm not in self.algorithms:
            raise ValueError(f"Algorithm {algorithm} not available")
        
        if self.swarm_state is None:
            self.initialize_swarm(bounds, fitness_function)
        
        optimizer = self.algorithms[algorithm]
        self._running = True
        
        try:
            for iteration in range(self.max_iterations):
                if not self._running:
                    break
                
                # Actualizar agentes
                if parallel:
                    self._update_agents_parallel(optimizer, fitness_function)
                else:
                    self._update_agents_sequential(optimizer, fitness_function)
                
                # Actualizar estado del enjambre
                self._update_swarm_state(iteration)
                
                # Guardar historial
                self.optimization_history.append({
                    'iteration': iteration,
                    'best_fitness': self.swarm_state.global_best_fitness,
                    'diversity': self.swarm_state.diversity_measure,
                    'algorithm': algorithm
                })
                
                # Verificar convergencia
                if self._check_convergence():
                    break
        
        finally:
            self._running = False
        
        return self._get_optimization_results()
    
    def collective_decision_making(self, 
                                  decision_space: List[np.ndarray],
                                  agent_preferences: Dict[str, np.ndarray]) -> Dict[str, Any]:
        """Toma de decisiones colectiva"""
        
        if not self.swarm_state:
            raise ValueError("Swarm not initialized")
        
        # Recopilar opiniones de agentes
        agent_opinions = []
        
        for agent in self.swarm_state.agents:
            if agent.id in agent_preferences:
                # Usar preferencias específicas del agente
                opinion = agent_preferences[agent.id]
            else:
                # Generar opinión basada en posición del agente
                opinion = self._generate_agent_opinion(agent, decision_space)
            
            agent_opinions.append(opinion)
        
        # Construir consenso
        consensus_result = self.consensus_builder.build_consensus(agent_opinions)
        
        # Métricas de calidad de decisión
        decision_quality = self._assess_decision_quality(
            consensus_result, agent_opinions
        )
        
        return {
            'decision': consensus_result['consensus'],
            'confidence': consensus_result['confidence'],
            'quality_metrics': decision_quality,
            'participation': len(agent_opinions),
            'algorithm_used': 'collective_consensus'
        }
    
    def adaptive_swarm_behavior(self, 
                               environment_feedback: Dict[str, float]) -> None:
        """Adapta comportamiento del enjambre según retroalimentación"""
        
        if not self.swarm_state:
            return
        
        # Analizar feedback del ambiente
        performance_score = environment_feedback.get('performance', 0.5)
        complexity_score = environment_feedback.get('complexity', 0.5)
        resource_availability = environment_feedback.get('resources', 1.0)
        
        # Adaptar roles de agentes
        self._adapt_agent_roles(performance_score, complexity_score)
        
        # Ajustar parámetros de algoritmos
        self._adapt_algorithm_parameters(performance_score, resource_availability)
        
        # Modificar estructura del enjambre si necesario
        if performance_score < 0.3:
            self._restructure_swarm()
    
    def get_swarm_analytics(self) -> Dict[str, Any]:
        """Obtiene analytics del enjambre"""
        if not self.swarm_state:
            return {"status": "Swarm not initialized"}
        
        # Estadísticas de agentes
        fitnesses = [agent.fitness for agent in self.swarm_state.agents]
        
        # Análisis de convergencia
        convergence_analysis = self._analyze_convergence()
        
        # Diversidad y exploración
        exploration_metrics = self._calculate_exploration_metrics()
        
        return {
            'swarm_size': len(self.swarm_state.agents),
            'current_iteration': self.swarm_state.iteration,
            'best_fitness': self.swarm_state.global_best_fitness,
            'average_fitness': np.mean(fitnesses),
            'fitness_std': np.std(fitnesses),
            'diversity_measure': self.swarm_state.diversity_measure,
            'convergence_rate': self.swarm_state.convergence_rate,
            'convergence_analysis': convergence_analysis,
            'exploration_metrics': exploration_metrics,
            'agent_roles': self._get_role_distribution()
        }
    
    # Métodos auxiliares privados
    def _assign_role(self, agent_index: int) -> str:
        """Asigna rol inicial a agente"""
        roles = ['explorer', 'exploiter', 'scout', 'coordinator']
        return roles[agent_index % len(roles)]
    
    def _calculate_diversity(self, agents: List[Agent]) -> float:
        """Calcula medida de diversidad del enjambre"""
        if len(agents) < 2:
            return 0.0
        
        positions = np.array([agent.position for agent in agents])
        center = np.mean(positions, axis=0)
        distances = [np.linalg.norm(pos - center) for pos in positions]
        
        return np.mean(distances)
    
    def _update_agents_parallel(self, 
                               optimizer: OptimizationAlgorithm,
                               fitness_function: Callable) -> None:
        """Actualiza agentes en paralelo"""
        with ThreadPoolExecutor(max_workers=4) as executor:
            futures = []
            
            for agent in self.swarm_state.agents:
                future = executor.submit(
                    self._update_single_agent, 
                    agent, optimizer, fitness_function
                )
                futures.append(future)
            
            # Recopilar resultados
            for i, future in enumerate(futures):
                self.swarm_state.agents[i] = future.result()
    
    def _update_agents_sequential(self, 
                                 optimizer: OptimizationAlgorithm,
                                 fitness_function: Callable) -> None:
        """Actualiza agentes secuencialmente"""
        for i, agent in enumerate(self.swarm_state.agents):
            self.swarm_state.agents[i] = self._update_single_agent(
                agent, optimizer, fitness_function
            )
    
    def _update_single_agent(self, 
                            agent: Agent,
                            optimizer: OptimizationAlgorithm,
                            fitness_function: Callable) -> Agent:
        """Actualiza un agente individual"""
        # Actualizar posición con algoritmo
        updated_agent = optimizer.update_agent(agent, self.swarm_state)
        
        # Evaluar nuevo fitness
        updated_agent.fitness = fitness_function(updated_agent.position)
        
        # Actualizar mejor personal
        if updated_agent.fitness > updated_agent.best_fitness:
            updated_agent.best_fitness = updated_agent.fitness
            updated_agent.best_position = updated_agent.position.copy()
        
        return updated_agent
    
    def _update_swarm_state(self, iteration: int) -> None:
        """Actualiza estado global del enjambre"""
        # Encontrar nuevo mejor global
        best_agent = max(self.swarm_state.agents, key=lambda a: a.fitness)
        
        if best_agent.fitness > self.swarm_state.global_best_fitness:
            self.swarm_state.global_best_fitness = best_agent.fitness
            self.swarm_state.global_best_position = best_agent.position.copy()
        
        # Actualizar métricas
        self.swarm_state.iteration = iteration
        self.swarm_state.diversity_measure = self._calculate_diversity(
            self.swarm_state.agents
        )
        self.swarm_state.convergence_rate = self._calculate_convergence_rate()
    
    def _calculate_convergence_rate(self) -> float:
        """Calcula tasa de convergencia"""
        if len(self.optimization_history) < 2:
            return 0.0
        
        recent_improvements = []
        for i in range(1, min(10, len(self.optimization_history))):
            current = self.optimization_history[-i]['best_fitness']
            previous = self.optimization_history[-i-1]['best_fitness']
            improvement = current - previous
            recent_improvements.append(improvement)
        
        return np.mean(recent_improvements) if recent_improvements else 0.0
    
    def _check_convergence(self, tolerance: float = 1e-6) -> bool:
        """Verifica convergencia del enjambre"""
        if len(self.optimization_history) < 10:
            return False
        
        recent_bests = [
            entry['best_fitness'] 
            for entry in self.optimization_history[-10:]
        ]
        
        return np.std(recent_bests) < tolerance
    
    def _get_optimization_results(self) -> Dict[str, Any]:
        """Obtiene resultados de optimización"""
        return {
            'best_position': self.swarm_state.global_best_position,
            'best_fitness': self.swarm_state.global_best_fitness,
            'iterations': self.swarm_state.iteration,
            'convergence_achieved': self._check_convergence(),
            'final_diversity': self.swarm_state.diversity_measure,
            'optimization_history': self.optimization_history
        }
    
    def _generate_agent_opinion(self, 
                               agent: Agent, 
                               decision_space: List[np.ndarray]) -> np.ndarray:
        """Genera opinión de agente para decisión colectiva"""
        # Selecciona opción más cercana a posición del agente
        distances = [
            np.linalg.norm(agent.position - option) 
            for option in decision_space
        ]
        
        # Añade ruido basado en rol del agente
        noise_factor = {
            'explorer': 0.3,
            'exploiter': 0.1,
            'scout': 0.5,
            'coordinator': 0.2
        }.get(agent.role, 0.2)
        
        best_option_idx = np.argmin(distances)
        opinion = decision_space[best_option_idx].copy()
        
        # Añadir ruido según rol
        noise = np.random.normal(0, noise_factor, len(opinion))
        opinion += noise
        
        return opinion
    
    def _assess_decision_quality(self, 
                                consensus_result: Dict[str, Any],
                                agent_opinions: List[np.ndarray]) -> Dict[str, float]:
        """Evalúa calidad de decisión colectiva"""
        if not agent_opinions:
            return {"quality_score": 0.0}
        
        consensus = consensus_result['consensus']
        
        # Métricas de calidad
        quality_metrics = {}
        
        # 1. Satisfacción promedio
        satisfactions = [
            1.0 / (1.0 + np.linalg.norm(opinion - consensus))
            for opinion in agent_opinions
        ]
        quality_metrics['average_satisfaction'] = np.mean(satisfactions)
        
        # 2. Equidad (distribución de satisfacciones)
        quality_metrics['fairness'] = 1.0 - np.std(satisfactions)
        
        # 3. Robustez (sensibilidad a outliers)
        outlier_opinions = [
            op for op in agent_opinions 
            if np.linalg.norm(op - consensus) > 2 * np.std([
                np.linalg.norm(o - consensus) for o in agent_opinions
            ])
        ]
        quality_metrics['robustness'] = 1.0 - len(outlier_opinions) / len(agent_opinions)
        
        # 4. Score compuesto
        quality_metrics['quality_score'] = (
            0.4 * quality_metrics['average_satisfaction'] +
            0.3 * quality_metrics['fairness'] +
            0.3 * quality_metrics['robustness']
        )
        
        return quality_metrics
    
    def _adapt_agent_roles(self, performance_score: float, complexity_score: float):
        """Adapta roles de agentes según performance"""
        if performance_score < 0.4:
            # Incrementar exploradores si performance es baja
            for agent in self.swarm_state.agents[:len(self.swarm_state.agents)//3]:
                agent.role = 'explorer'
        elif performance_score > 0.8:
            # Incrementar explotadores si performance es alta
            for agent in self.swarm_state.agents[:len(self.swarm_state.agents)//2]:
                agent.role = 'exploiter'
    
    def _adapt_algorithm_parameters(self, performance_score: float, resource_availability: float):
        """Adapta parámetros de algoritmos"""
        # Adaptar PSO
        if 'pso' in self.algorithms:
            pso = self.algorithms['pso']
            if performance_score < 0.5:
                pso.w *= 1.1  # Incrementar exploración
                pso.c1 *= 0.9  # Reducir componente cognitivo
            else:
                pso.w *= 0.95  # Reducir exploración
                pso.c2 *= 1.05  # Incrementar componente social
    
    def _restructure_swarm(self):
        """Reestructura enjambre si performance es muy baja"""
        # Reinicializar 30% de agentes con peor performance
        worst_agents = sorted(
            self.swarm_state.agents, 
            key=lambda a: a.fitness
        )[:len(self.swarm_state.agents)//3]
        
        for agent in worst_agents:
            agent.position = np.random.uniform(-10, 10, self.dimensions)
            agent.velocity = np.random.uniform(-1, 1, self.dimensions)
            agent.fitness = float('-inf')
    
    def _analyze_convergence(self) -> Dict[str, Any]:
        """Analiza patrones de convergencia"""
        if len(self.optimization_history) < 5:
            return {"status": "Insufficient data"}
        
        # Analizar tendencias
        fitness_trend = [entry['best_fitness'] for entry in self.optimization_history[-10:]]
        diversity_trend = [entry['diversity'] for entry in self.optimization_history[-10:]]
        
        return {
            'fitness_improvement_rate': np.mean(np.diff(fitness_trend)),
            'diversity_trend': 'decreasing' if np.mean(np.diff(diversity_trend)) < 0 else 'stable',
            'convergence_speed': 'fast' if abs(np.mean(np.diff(fitness_trend))) > 0.1 else 'slow',
            'premature_convergence_risk': 'high' if np.mean(diversity_trend) < 0.1 else 'low'
        }
    
    def _calculate_exploration_metrics(self) -> Dict[str, float]:
        """Calcula métricas de exploración"""
        if not self.swarm_state:
            return {}
        
        positions = np.array([agent.position for agent in self.swarm_state.agents])
        
        # Cobertura del espacio de búsqueda
        ranges = np.max(positions, axis=0) - np.min(positions, axis=0)
        space_coverage = np.mean(ranges)
        
        # Distribución de agentes
        center = np.mean(positions, axis=0)
        distances_to_center = [np.linalg.norm(pos - center) for pos in positions]
        distribution_uniformity = 1.0 - np.std(distances_to_center) / np.mean(distances_to_center)
        
        return {
            'space_coverage': space_coverage,
            'distribution_uniformity': distribution_uniformity,
            'exploration_pressure': np.mean([
                1.0 if agent.role in ['explorer', 'scout'] else 0.0 
                for agent in self.swarm_state.agents
            ])
        }
    
    def _get_role_distribution(self) -> Dict[str, int]:
        """Obtiene distribución de roles en el enjambre"""
        if not self.swarm_state:
            return {}
        
        roles = {}
        for agent in self.swarm_state.agents:
            roles[agent.role] = roles.get(agent.role, 0) + 1
        
        return roles
    
    def stop_optimization(self):
        """Detiene optimización en curso"""
        self._running = False