IMPROVE ACCURACY POSE ESTIMATION OF PALLETS USING INTEGRATION OF CAMERA SENSOR BASED ON ARUCO AND LASER RANGEFINDER WITH EXTENDED KALMAN FILTER AND PARTICLE FILTER ALGORITHMS

• Main Author: Fijar Aswad, Muhammad
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/79453
• Institution: Institut Teknologi Bandung

Due to the increasing demands for efficiency, accuracy, and safety in warehouse management systems, the implementation of autonomous forklifts has emerged as a viable solution. A crucial stage in deploying autonomous forklifts involves pallet detection and pose estimation. Pallet detection refers to identifying and determining the location of pallets, while pose estimation involves utilizing information obtained from pallet recognition to identify the position and orientation of the pallet. In this research, the authors propose an approach to detect and estimate pallet pose using a camera with ArUco markers and a laser rangefinder sensor. These two sensors are fused through the Extended Kalman Filter (EKF) and Particle Filter (PF) algorithms. The pallet detection process using ArUco markers achieved an average detection speed of 28 frames per second in real-time at a measuring range of 30-320 cm. The pallet pose estimation process using the integrasion method with the EKF algorithm demonstrated an accuracy improvement of 60,75% and 31,78% for distance estimation over individual camera and laser measurements, respectively. Additionally, the EKF algorithm showed a 30,56% and 41,35% improvement in angle estimation over the camera and laser, respectively. Meanwhile, the PF algorithm provided an accuracy improvement of 62,30% and 47,99% for distance estimation over the camera and laser, with angle estimation improvements of 23,81% and 39,81%. The EKF algorithm exhibited a precision improvement of 62,79% and 5,88% for distance estimation over the camera and laser, respectively. For angle estimation, precision improved by 21,04% and 40,63% over the camera and laser. The PF algorithm demonstrated precision improvements in distance estimation of 77,78% and 0% and angle estimation of 27,74% and 44,55% over the camera and laser, respectively. In conclusion, the camera and laser rangefinder integrasion method with EKF and PF algorithms yields higher accuracy and precision than individual sensor measurements.