Electronic Theses and Dissertation

Simpang bersinyal merupakan elemen penting dalam sistem transportasi perkotaan
yang kapasitasnya dipengaruhi oleh arus jenuh dasar (Base Saturation Flow
Rate/BSFR). Namun, model arus jenuh pada PKJI 2023 masih mengadopsi
pendekatan lalu lintas homogen sehingga belum mampu merepresentasikan kondisi
lalu lintas perkotaan di Indonesia yang heterogen, didominasi sepeda motor, dan
dipengaruhi Weak Lane Discipline (WLD). Kondisi tersebut berpotensi
menimbulkan bias dalam estimasi kapasitas dan kinerja simpang bersinyal.
Penelitian ini bertujuan mengembangkan model estimasi BSFR berbasis
karakteristik lalu lintas lokal menggunakan pendekatan regresi linier Ordinary
Least Squares (OLS) dan Bayesian Markov Chain Monte Carlo (MCMC) dengan
Gibbs Sampling. Data penelitian diperoleh dari enam simpang bersinyal berlengan
empat di Kota Banda Aceh melalui survei jam puncak menggunakan Unmanned
Aerial Vehicle (UAV). Arus lalu lintas dikonversi ke dalam satuan mobil
penumpang (smp) sesuai PKJI 2023 untuk memperoleh karakteristik BSFR. Hasil
penelitian menunjukkan bahwa lebar pendekat berpengaruh signifikan terhadap
BSFR, sedangkan waktu hilang awal berpengaruh negatif. Model Bayesian
menghasilkan formulasi terbaik sebesar 403Le dengan tingkat akurasi lebih tinggi
dibandingkan model OLS sebesar 350Le maupun model PKJI sebesar 600Le yang
cenderung mengestimasi kapasitas lebih tinggi dari kondisi aktual. Nilai RMSEA
model Bayesian sebesar 8,638% menunjukkan kinerja yang jauh lebih baik
dibandingkan PKJI sebesar 51%. Dominasi sepeda motor memperkuat pengaruh
WLD yang menyebabkan fluktuasi arus pada awal fase hijau dan mempengaruhi
pelepasan arus jenuh. Penelitian ini menegaskan bahwa pendekatan probabilistik
berbasis data lokal lebih mampu memodelkan BSFR pada kondisi lalu lintas
heterogen, serta berpotensi menjadi dasar pengembangan model kapasitas simpang
yang lebih adaptif di Indonesia

Signalized intersections are important elements of urban transportation systems whose capacity is influenced by the Base Saturation Flow Rate (BSFR). However, the saturation flow model in PKJI 2023 still adopts a homogeneous traffic approach, which has not been able to adequately represent the heterogeneous urban traffic conditions in Indonesia, characterized by motorcycle dominance and influenced by Weak Lane Discipline (WLD). These conditions may lead to bias in estimating the capacity and performance of signalized intersections. This study aims to develop a BSFR estimation model based on local traffic characteristics using Ordinary Least Squares (OLS) linear regression and Bayesian Markov Chain Monte Carlo (MCMC) with Gibbs Sampling approaches. The research data were collected from six four-legged signalized intersections in Banda Aceh City through peak-hour surveys using Unmanned Aerial Vehicle (UAV). Traffic flows were converted into passenger car units (pcu) according to PKJI 2023 to obtain BSFR characteristics. The results indicate that the effective approach width significantly affects BSFR, while start-up lost time has a negative effect. The Bayesian model produced the best formulation of 403Le, with a higher level of accuracy compared to the OLS model of 350Le and the PKJI model of 600Le, which tended to overestimate the actual intersection capacity. The RMSEA value of the Bayesian model was 8.638%, demonstrating substantially better performance than the PKJI model at 51%. Motorcycle dominance intensified the influence of WLD, causing traffic flow fluctuations during the early green phase and affecting saturation flow discharge. This study confirms that a probabilistic approach based on local traffic data is more capable of modeling BSFR under heterogeneous traffic conditions and has the potential to serve as a basis for developing a more adaptive intersection capacity model in Indonesia