Reduced capacity on motorways can easily lead to significant congestion. This congestion is a major contributor to environmental pollution, harming the livability of the peri-urban environment and public health. In this study, we have addressed the congestion caused by lane closures on motorways, one of the many difficulties encountered in the lane closure problem. To overcome this problem, the so-called variable speed limit control, a traffic management system is a helpful tool that improves overall traffic flow characteristics - travel time, waiting time, and queue length - and reduces critical sustainability indicators such as fuel consumption and CO2 and NOx emissions. Deep Learning has repeatedly been shown to be an excellent solution to this problem. Hence, this study aims to use Reinforcement Learning to address the traffic management system and to find a general solution to congestion caused by the reduction of highway capacity to apply the model regardless of the number of lanes, improving and surpassing the results achieved in the literature in several aspects.