SIMULASI PENGATURAN LAMPU LALU LINTAS DENGAN DETEKSI ARUS KENDARAAN MENGGUNAKAN YOLO

Nunik Pratiwi; Muhamad Arief Rachman Hakim;

ABSTRAK

Simulasi Pengaturan Lampu Lalu Lintas Dengan Deteksi Arus

Kendaraan Menggunakan YOLO
Muhamad Arief Rachman Hakim

Ketidakseimbangan antara volume kendaraan dan kapasitas infrastruktur sering diperburuk oleh sistem lampu
lalu lintas statis yang tidak responsif. Penelitian ini mengembangkan sistem pengaturan lampu lalu lintas adaptif
berbasis Computer Vision menggunakan algoritma YOLOv8s yang terintegrasi dengan logika distribusi waktu
proporsional. Model deteksi yang dilatih menunjukkan performa andal dengan nilai mAP@0,5 sebesar 0,883.
Hasil implementasi menunjukkan variabilitas kinerja berdasarkan kondisi pencahayaan; sistem mencapai
akurasi perhitungan (Mean Absolute Percentage Accuracy/MAPA) sebesar 80,97% pada siang hari. Meskipun
akurasi terkoreksi menjadi 61,53% pada malam hari akibat tantangan visual, sistem terbukti memiliki
ketahanan (robustness) tinggi karena logika proporsional tetap mampu menjaga rasio kepadatan antarjalur
secara presisi. Sistem berhasil menghilangkan inefisiensi waktu (wasted green time) dengan menghasilkan
durasi adaptif pada rentang 8–45 detik, serta menjamin keamanan operasional melalui mekanisme fail-safe
batas minimum lampu hijau 5 detik. Penelitian ini membuktikan bahwa sistem mampu menciptakan distribusi
sinyal yang adil dan adaptif terhadap dinamika lalu lintas nyata.
Kata Kunci: Deteksi Kendaraan; Lampu Lalu Lintas Adaptif; Metode Proporsional; Visi Komputer; YOLOv8

ABSTRACT

Simulation of Traffic Light Control with Vehicle Flow Detection

Using YOLO
Muhamad Arief Rachman Hakim

Traffic congestion is often exacerbated by fixed-time traffic light systems that fail to respond to dynamic vehicle
density. This study develops an adaptive traffic light control system based on Computer Vision using the
YOLOv8s algorithm integrated with proportional time distribution method. The trained detection model
demonstrated reliable performance with an mAP@0,5 of 0,883. Implementation results showed performance
variability based on lighting conditions; the system achieved a calculation accuracy (Mean Absolute
Percentage Accuracy/MAPA) of 80,97% during the day. Although accuracy decreased to 61,53% at night due
to visual challenges, the system proved to be robust as the proportional method successfully maintained
accurate density ratios between lanes. The system succeeded in eliminating wasted green time by generating
adaptive durations ranging from 8–45 seconds and ensuring operational safety through a 5-second green light
minimum fail-safe mechanism. This study confirms that the proposed system can create a fair and adaptive
signal distribution for real-world traffic dynamics.
Keywords: Adaptive Traffic Light; Computer Vision; Proportional Method; Vehicle Detection; YOLOv8