Copyright © 2018 Authors. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. International Journal of Engineering & Technology, 7 (4.36) (2018) 404-408 International Journal of Engineering & Technology Website: www.sciencepubco.com/index.php/IJET Research paper Energy and Path Optimization of Robot Arm Simulator Via Multi-Objective Evolutionary Algorithm Z. Mohamed 1,2 *, N.S. Khusaini 2,3 , M.A.M Anuar 1,2 , R. Ramly 1 , M.A Anuar 1,2 , K.S.M Sahari 4 , 1 Faculty of Mechanical Engineering Universiti Teknologi MARA, Shah Alam Selangor, Malaysia 2 Sports Engineering & Artificial Intelligent Center, Faculty of Mechanical Engineering Universiti Teknologi MARA, Shah Alam Selangor, Malaysia 3 Faculty of Mechanical Engineering, Universiti Teknologi MARA, Kampus Permatang Pauh, Pulau Pinang, Malaysia 4 Department of Mechanical Engineering, Universiti Tenaga Nasional, Kajang Selangor, Malaysia *Corresponding author E-mail: zulkifli127@salam.uitm.edu.my Abstract The performance of robot arm motion generated via neural network are presented in this paper. The robot arm motion for obstacle avoid- ance simultaneously optimizing three functions; minimum distance, minimum time and minimum energy are generated. Four different initial and goal position had been chosen to test and analyze the performance of generated neural controller. The same neural controllers can be employed to a different range of initial and goal position. The motion generated yield good results in the simulator. In this re- search a new approach for intelligent robot arm path and motion generation are successfully implemented. Keywords: Robot arm, Genetic algorithm, Neural network, Multi-objective evolutionary algorithm. 1. Introduction Human being is said to have unique traits compared to other living creatures. Human being has the ability to think constructively, and fulfill daily needs by performing various kind of task, which may be complicated and need further scrutiny. However, as time goes on, smart humanoid robots are developed to do simple daily rou- tines which are normally done by humans. For instance, taking order in the restaurant and as a receptionist in the hotel. As the technology progresses, researchers are looking into possi- bilities to create humanoid robots with the capability and ability to mimic simple human movement. The movement is related to reaching and moving an object from one place to another. As easy as it may seem, there are certain requirements that should be taken into consideration. For instance, if the humanoid robot is pro- grammed to perform simple movement such as to move an object, it needs to identify at least seven different elements namely shape, size, position, color (for object’s identification purposes), as well as optimum distance, speed and energy (for the robot’s hand movement). This is crucial to ensure smooth and accurate travel- ling path, especially, if there is a lot of obstacles around the cho- sen or specific object. Hence, there is a need to clearly determine the object’s position, the robot’s hand initial position (starting point), as well as the travelling path. To ensure a successful implementation, motion generation charac- teristics such as optimum speed and distance, as well as obstacle avoidance should be clearly determined for the robot’s hand to mimic or perform as how human’s hand performed. Motion gen- eration characteristics is the combination of two or more motion characteristics which will affect the robot’s performance, depend- ing on the requirements which is stated by the researcher. The approach to the development of motion generation character- istics has been conducted in previous research using two different methods namely Rapid-Exploring Random Trees (RRT) and Rap- id-Exploring Dense Trees (RDT). For RRT algorithm, the main focus is to optimize the degrees of freedom when the robotic arm named ARMAR-III is performing the given task [1]–[4]. On the other hand, RDT is related to fine-tuning the parameters to detect crash automatically until the final or best solution is obtained [3]. The finding has proven that the performance for RDT is superior than RRT. In other works, [5] utilize a segmented positioning method and it had been implemented in an established service robot manipulator motion. Higher accuracy had been achieved by utilizing the robot vision system. A minimum time industrial robot arm motion gen- eration had been proposed by [6]. In this study, the robot manipu- lators’ dynamic model is utilized to regulate the maximum kine- matic constraints. On the other hand, a single objective robot arm motion had been introduced by [7]. However, the minimum dis- tance, minimum velocity and minimum acceleration objective function for robot arm motion generation had been implemented as separate entity. Opportunely, a motion generation via multi-objective characteris- tics has been proposed by [8] and [9]. Five objective functions are considered for the study. The five objective functions are con- structed to achieve minimum angular displacement, minimum, Cartesian distance, minimum angular velocity, minimum Carte- sian velocity, as well as minimum energy. A planar robot is used to test the performance and capability of the proposed method. A similar approach is applied by [10] for generating parallel kine- matics machine motion. Three different objective functions were chosen in their work; namely minimum shaking force, minimum