mo_pso_rs's Introduction

mo_pso_rs

An implementation of a multi-objective particle swarm optimization algorithm utilizing methods proposed in Strategies for finding good local guides in multi-objective particle swarm optimization (MOPSO).

This crate is mostly a Rust rewrite of the implementation of MOPSO in the PSOTS toolbox with additional support for constrained optimization with arbitrary constraints.

Currently, only two objectives are supported, but the implementation should be easily extendable to more objectives.

Constraints

This implementation supports arbitrary constraints which are handled using a "Preserving feasibility technique" as described in the paper Constraint-Handling Techniques for Particle Swarm Optimization Algorithms.

Example

use mo_pso_rs::{MoPso, MoPsoConfig, Particle, Problem};

struct TestProblem {
    radius: f64
}
impl Problem for TestProblem {
    fn objective(&self, p: &[f64]) -> [f64; 2] {
        [p[0].powi(2), -p[1].powi(2)]
    }

    fn is_feasible(&self, p: &[f64]) -> bool {
        // Only allow point outside circle with radius self.radius
        p[0].powi(2) + p[1].powi(2) >= self.radius.powi(2)
    }
}


fn main() {
    let problem = TestProblem { radius: 2.0 };
    
    let config = MoPsoConfig {
        max_iter: 100,
        c_1: 0.4, 
        c_2: 0.4,
        w_start: 0.9,
        w_end: 0.4,
        archive_size: 100,
    };

    let mut particles = Vec::new();
    for _ in 0..100 {
        particles.push(Particle::random(2));
    }

    let optimizer = MoPso::new(config, &problem, particles);
    let pareto_front = optimizer.optimize();

    for particle in pareto_front {
        println!("{:?}", particle.position);
    }
}

Recommend Projects