PROTOTIPE HUMAN FOLLOWING ROBOT
MENGGUNAKAN TENSOR FLOW LITE
PADA RASPBERRY PI
Total View This Week0
Total View This Week0
Institusion
INSTITUT TEKNOLOGI DIRGANTARA ADISUTJIPTO
Author
BAGUS ARYANDIKA, AHNAF
Subject
T Technology (General)
Datestamp
2024-05-06 02:47:18
Abstract :
In this research, a prototype human following robot was designed and implemented using Tensor Flow Lite on Raspberry Pi hardware where the robot is equipped with the ability to follow human movements to help aircraft technicians carry aircraft maintenance tool kits. The movement of the robot is based on input from the camera sensor that follows the movement of the object through determining the coordinates of the midpoint of the camera frame. The input image from the sensor is processed by the YOLO v3-Tiny deep learning model with the stages of determining the bounding box, calculating Euclidean Distance, and determining the pixel value of the camera as the basis for the robots movement. If the Euclidean Distance value is less than 70, then the robot will stop. If it is the other way around, then the robot will move. The movement of the robot is regulated by the L298N motor driver through two Direct Current (DC) motors for the left wheel and right wheel. YOLO v3-Tiny is applied using the Tensor Flow Lite machine learning library on the Raspberry Pi. The results showed that at a distance of 2 m, the average robot movement speed is 0,18 m/second, and the travel time is 11 seconds with the use of voltage of 6,5 V, and light intensity of more than 250 lux. The minimum and maximum distance of object detection is as far as 120 cm and 400 cm with a motor turning accuracy of 79,2% and object detection accuracy of 95,8%. The robot is capable of carrying a maximum load weighing 600 g.
In this research, a prototype human following robot was designed and implemented using Tensor Flow Lite on Raspberry Pi hardware where the robot is equipped with the ability to follow human movements to help aircraft technicians carry aircraft maintenance tool kits. The movement of the robot is based on input from the camera sensor that follows the movement of the object through determining the coordinates of the midpoint of the camera frame. The input image from the sensor is processed by the YOLO v3-Tiny deep learning model with the stages of determining the bounding box, calculating Euclidean Distance, and determining the pixel value of the camera as the basis for the robots movement. If the Euclidean Distance value is less than 70, then the robot will stop. If it is the other way around, then the robot will move. The movement of the robot is regulated by the L298N motor driver through two Direct Current (DC) motors for the left wheel and right wheel. YOLO v3-Tiny is applied using the Tensor Flow Lite machine learning library on the Raspberry Pi. The results showed that at a distance of 2 m, the average robot movement speed is 0,18 m/second, and the travel time is 11 seconds with the use of voltage of 6,5 V, and light intensity of more than 250 lux. The minimum and maximum distance of object detection is as far as 120 cm and 400 cm with a motor turning accuracy of 79,2% and object detection accuracy of 95,8%. The robot is capable of carrying a maximum load weighing 600 g.
Download
book
BibTex
Latex, Jabref
cloud_download
Original Resource
url resource Institution
