Implementation of Augmented Multiobjective Particle Swarm Optimization algorithm (MOPSO hybridized with local search).
- Clone the project to a directory of choice using
git clone https://github.com/siddarthgopalakrishnan/hybrid-mopso.git - Change into the project directory
- Create a virtual environment by running
python3 -m venv <name_of_venv> - Activate the venv using
source <name_of_venv>/bin/activate - Install the required libraries
- Run the code by running
python3 main.py
- Binh and Korn (2 input dimensions, 2 output dimensions)
- Tanaka (2 input dimensions, 2 output dimensions)
- Osyczka (6 input dimensions, 2 output dimensions)
- Zitzler-Deb-Thiele (ZDT2) (30 input dimensions, 2 output dimensions)
The algorithm was tested only on two-objective functions, although it works for many objective functions as well, with an obvious increase in the computational complexity.
- Running times (in seconds) of the algorithms for the above test functions
| Algorithm | Binh |
Tanaka |
Osyczka |
ZDT2 |
|---|---|---|---|---|
| MOPSO+ | 107 | 51 | 76.75 | 127.4 |
| SMPSO | 3.6 | 0.72 | 0.97 | 3.3 |
| NSGA2 | 1.3 | 1.21 | 1.22 | 1.38 |
-
For comparing the solutions of the algorithm, we use the comparison method suggested by the 4th reference paper where they see the contribution of each algorithm. To give a brief explanation of the method, we take the pareto fronts of algorithms A and B, and see how many solutions in the combined pareto front are contributed by algorithms A and B.
- Nt: Total elements present in each pareto front
- Nc: Common solutions that have been contributed by both algorithms
- Nu: Unique solutions contributed by each algorithm
- Nr: Number of solutions removed during the process of merging the pareto fronts
-
With these statistics, we can observe the Nu and Nc and comment on the performance of both the algorithms.
| Function | MOPSO+ Pareto front | Comparison Metric |
|---|---|---|
Binh |
 |
 |
Tanaka |
 |
 |
Osyczka |
 |
 |
ZDT2 |
 |
 |
- Discrete search spaces
- EPANET testing (resilience and diameter) (eg: Hanoi network, Two-loop network)
- Optimize non-dominated sort (currently O(M*N^2))
-
Papers
- MOPSO: a proposal for multiple objective particle swarm optimization by C.A. Coello Coello. Paper link
- Handling multiple objectives with particle swarm optimization by C.A. Coello Coello. Paper Link
- An Improved Multiobjective Particle Swarm Optimization Based on Culture Algorithms by Chunhua Jia and Hong Zhu. Paper Link
- Water Distribution System Design Using Multi-Objective Particle Swarm Optimisation by Mahesh B. Patil and A. Vasan. Paper Link
-
Github repositories
-
Resources and websites