smart-pch-tsp-rs

• Universal Improver — Takes ANY path — returns a shorter one
• Generic Points — Works with any type implementing the Point trait
• Statistical Path Synthesis — Uses transition frequency analysis
• Parallel Processing — Multi-agent system for faster computation
• Better Quality — Improves paths by 1-2% on average (up to 3% on some datasets)
• No Dependencies — Only uses Rust standard library

PCH improves paths through statistical analysis:

1. Accept ANY initial path (from any algorithm)
2. Build candidate framework (top N candidates per position)
3. Generate hypotheses with random permutations
4. Collect transition statistics (weighted by improvement)
5. Synthesize better path from maximum-weight transitions

Add to your Cargo.toml:

[dependencies]
smart-pch-tsp = "0.1"

use smart_pch_tsp::{City, pch_improve};

let points = vec![
    City { x: 0.0, y: 0.0 },
    City { x: 1.0, y: 0.0 },
    City { x: 0.0, y: 1.0 },
];

let initial_path = vec![0, 1, 2, 0];
let initial_distance = 4.0;

let (improved_distance, improved_path) = pch_improve(
    &points,
    &initial_path,
    initial_distance,
    50,    // permutations_per_hypothesis
    1000,  // batch_size
    30,    // time_limit (seconds)
    300,   // top_candidates
    12,    // num_agents
);

println!("Improved: {:.2} -> {:.2}", initial_distance, improved_distance);

This library is part of the NP Problem Ecosystem — a comprehensive suite of exact and heuristic solvers for the Traveling Salesman Problem:

All projects are grounded in the Position-Candidate-Hypothesis (PCH) paradigm for NP-complete problems.

# Clone repository
git clone https://github.com/smartlegionlab/smart-pch-tsp-rs
cd smart-pch-tsp

# Build
cargo build

# Run tests
cargo test

# Run documentation tests (checks code examples in docs)
cargo test --doc

# Run examples
cargo run --example basic --release
cargo run --example improve_dg --release

# Run benchmarks
cargo bench

# Build documentation
cargo doc --open

By using this software, you agree to the full disclaimer terms.

Software provided "AS IS" without warranty. You assume all risks.

Full legal disclaimer: See DISCLAIMER.md

License: BSD 3-Clause License