Self-Organizing Neuroevolution for Solving Carpool Service Problem With Dynamic Capacity to Alternate Matches

Area of Science:

• Artificial Intelligence
• Operations Research
• Transportation Science

Background:

• Traffic congestion poses environmental challenges, with carpooling identified as a key mitigation strategy.
• Intelligent carpool systems, accessible via mobile apps, require optimized resource distribution, known as the carpool service problem (CSP).
• Previous CSP solutions relied on exact or pure metaheuristic optimization, with evolutionary computation showing promise.

Purpose of the Study:

• To propose a novel neuroevolutionary framework for solving the carpool service problem (CSP).
• To introduce the self-organizing map-based neuroevolution (SOMNE) solver, integrating neural learning and evolutionary mechanisms.
• To demonstrate the superiority of the SOMNE solver over existing approaches for CSP optimization.

Main Methods:

• Employing a neuroevolution framework to develop the self-organizing map-based neuroevolution (SOMNE) solver.
• Utilizing a self-organizing map (SOM)-like network to represent abstract CSP solutions.
• Training the network through a combination of neural learning and evolutionary computation.

Main Results:

• The SOMNE solver achieved superior results compared to previous state-of-the-art approaches for the carpool service problem.
• Demonstrated significant improvements in optimizing the primary objective functions of the CSP.
• Visualizations confirmed the effectiveness and efficiency of the evolutionary neural learning process.

Conclusions:

• Neuroevolution, specifically the SOMNE solver, offers a powerful and effective approach to the carpool service problem.
• The proposed method enhances the intelligent distribution of carpool participant resources.
• SOMNE provides a promising advancement over traditional optimization techniques in intelligent transportation systems.