Robot Navigation in Palm Tree Cluster Areas: Overcoming Challenges through a New Vision-Ultrasonic Fusion System

Abstract

Ensuring the upkeep of a large public palm tree cluster area can be time-consuming and resource-intensive. This is particularly true with challenges such as the frequent falling of dates and palm fronds, which require regular and efficient cleaning and maintenance. With the recent advancements in computing capabilities, robots have been efficiently utilized for outdoor cleaning applications. However, navigating autonomous robots in areas of palm tree clusters presents several challenges stemming from the unique characteristics of such environments. One significant challenge is the weakening of GPS signals due to occlusion by dense foliage, which hinders accurate localization. Additionally, the fluctuating and limited sunshine, caused by the dense canopy of palm trees, introduces variability in lighting conditions, which complicates visual recognition tasks. Moreover, the challenge of discerning trimmed palm tree trunks from similar objects such as light poles adds another layer of complexity to the navigation process. In this paper, a robust outdoor navigation system that addresses these challenges is proposed by fusing vision and ultrasonic data to ensure enhanced localization. Leveraging YOLO-v4 for object detection, the proposed system ensures reliable identification of palm tree trunks, light poles, and other landmarks crucial for navigation. After the landmark detection, the proposed system utilizes the ultrasonic data to find the distance to the recognized object and then adjusts the location of the robot based on the readings. The proposed navigation system was tested experimentally in an area of dense palm tree clusters at the University of Sharjah. The experimental validation has proven the effectiveness of the proposed approach in facilitating precise and efficient robot navigation within the palm tree cluster area, overcoming the limitations imposed by weak GPS signals, lighting variations, and visual ambiguities arising from trimming practices.