Citation

Tan, Chi Jie and Munjer, M.A. and Hayashi, Eiji and Lim, Way Soong and Mowshowitz, Abbe Evolutionary Behavior Tree Synthesis for MultiObjective Autonomous Beach Cleaning Robot Using Genetic Programming with Behavioral Elitism. Journal of Robotics and Control (JRC), 7 (3). pp. 3828-3847. ISSN 27155056

Text 29205-Article Text-125036-1-10-20260811.pdf - Published Version Restricted to Repository staff only Download (1MB)

Abstract

This paper presents a genetic programming framework for automatically synthesizing and evolving behavior trees for autonomous robots operating in complex beach cleanup scenarios. The framework enables robots to address multi-objective tasks that require balancing garbage collection, human interaction, battery management, capacity control, and task completion efficiency. To support this, the proposed approach combines staged fitness evaluation with adaptive reward modulation and Niche Protection, allowing distinct behavioral competencies to emerge without premature convergence. A key contribution is the introduction of Behavioral Elitism, an evolution strategy that preserves highperforming and behaviorally diverse individuals to prevent premature convergence. The system incorporates multi-epoch evolutionary training using tournament selection, adaptive crossover, and context-aware weighted mutation operators to sustain behavioral diversity and promote the emergence of integrated multi-objective competencies. Experimental validation in a high-fidelity simulated beach environment demonstrates substantial performance gains over a baseline genetic programming system. The proposed method achieved a 42.6% improvement in maximum fitness and a 67.6% enhancement in behavioral diversity, with large effect sizes (Cohen’s d = 1.49–5.03) indicating strong practical significance. The evolved robots exhibited complex, emergent behaviors, including efficient garbage collection, adaptive battery-aware task scheduling, dynamic human-robot cooperation, and sustained operational balance. These results confirm that maintaining behavioral diversity is essential for achieving robust multi-objective coordination in autonomous systems. The proposed framework establishes a scalable foundation based on objective-agnostic niche definitions, population-driven adaptive reward amplification, and modular behavior tree representation. It is designed for future adaptive robotic applications in real-world multi-task environments.

Item Type:	Article
Uncontrolled Keywords:	Evolutionary Robotics, Multi-Objective Optimization, Autonomous Systems,
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7800-8360 Electronics > TK7885-7895 Computer engineering. Computer hardware
Divisions:	Faculty of Engineering and Technology (FET)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	02 Oct 2026 04:21
Last Modified:	02 Oct 2026 04:21
URII:	http://shdl.mmu.edu.my/id/eprint/16833

Downloads

Edit (login required)