Improving Solution Quality for Experience-Based Framework Through Clustering Algorithms

Abstract

This paper presents an extension for the current developed experience-based frameworks. The current experience-based scheme depends on executing two parallel threads; one tries to solve the problem using traditional approaches, while the other thread uses experience from past solutions for solving it. Once one of these threads solves the problem, its solution is promoted to be the problem's solution, yet disregarding the solution quality. However, our extension to experience-based frameworks uses clustering to decides which thread will solve the problem while maintaining solution quality. We have used experience-based motion planners for benchmarking our approach, where the presented results demonstrate that this approach works with different experience representations while maintaining better path quality, experience utilization, and reduced computational cost.

Keywords

• Artificial intelligence
• case-based reasoning
• experience-based algorithms
• motion planning
• sampling-based algorithms