Development and Empirical Evaluation of a Biomimetic Autonomous Robotic Arm for Manipulating Objects with Diverse geometries

Abstract

Abstract This paper discusses the design and development of a biomimetic robotic arm, elaborating on the experiments conducted with the developed arm to handle objects of diverse geometries, as well as evaluating its agility during grasping tasks. When automating fruit harvesting, it is crucial to minimize damage to leaves, as they play an essential role in the photosynthesis process. Thus, a versatile prehensile design is imperative for grasping fruits with various shapes. Existing technologies for harvesting fruit meant for processing are limited to soft, fresh fruit due to the risk of mechanical damage. As an alternative, a robotic system that emulates human fruit picking can improve fruit quality while maintaining efficiency. Consequently, a robotic hand with deformable fingers inspired by the human arm is developed. The robotic system must also be cost-effective. A single-gear motor is utilized to control the arm’s functions and ensure agile responsiveness when grasping objects with different shapes, incorporating a self-adaptive mechanism. During the development process, several grasping tests are conducted to evaluate the arm’s ability to handle basic shape primitives such as spheres and cylinders. The goal is to offer an alternative to manual fruit picking by creating a system capable of identifying, locating, and detaching fruit without causing damage to the fruit or tree. The robot is also equipped with A. In technology such as object detection and manipulation, the model is trained using a convolutional neural network for grasping the objects with appropriate pressures.