How does the GOLC model improve traffic safety and flow?

The Global Optimal Lane-Changing (GOLC) model was designed to make lane‑change decisions by considering not only the immediate effect on neighboring vehicles but also the impact on the entire affected platoon. Using a kinematic wave, the model predicts how a lane change will alter speed and spacing for all vehicles downstream, reducing harsh braking and unnecessary maneuvers. Tested in the SUMO simulator, GOLC was benchmarked against MOBIL and LC2013. In medium‑ to high‑flow three‑lane scenarios, GOLC cut average delay by 3.4% to 46.8% compared to MOBIL. It also reduced unnecessary lane changes by a factor of 1.1 versus LC2013, improving safety. During incident conditions, the model sustains higher speeds and lower travel times while limiting speed dispersion and sharp decelerations. These findings show that embedding macroscopic traffic theory into microscopic driving choices can yield smarter driver‑assistance systems and safer roads for everyone.

The GOLC model is grounded in an optimization framework that minimizes a global cost function comprising delay, energy consumption, and collision risk terms. The cost function incorporates the Lighthill‑Whitham‑Richards (LWR) kinematic wave equations to describe traffic density as a propagating wave, enabling the calculation of the spatiotemporal impact of a lane change on the entire downstream segment. Contrary to local models such as MOBIL, which evaluate only the speed gain relative to the lane‑changing vehicle and its immediate neighbors, GOLC accounts for the cumulative effect on all vehicles within the affected platoon, solving a constrained optimization problem via dynamic programming. SUMO simulations revealed that, under flow rates of 1500–2000 vehicles per hour per lane, GOLC reduces speed dispersion by 22% compared to LC2013 and cuts median deceleration by 18%. Moreover, the number of lane changes triggered by external disturbances (e.g., emergency vehicles) is lower by 15%, indicating greater robustness under non‑stationary traffic conditions. These results suggest that embedding macroscopic traffic models into microscopic decision‑making processes can be implemented on‑board via short‑term prediction algorithms, paving the way for advanced ADAS features such as cooperative lane changing and lane‑keeping assistance in complex urban environments.