Due to the increasing vehicle population and situation awareness in urban traffic systems, the transportation of people and goods is becoming more complex, which will soon warrant advanced and adaptive solutions to manage those systems effectively. In this research, we present a Hybrid Deep Reinforcement Learning (HDRL) framework using a method to manage traffic signals, route vehicles away from congestion, and reduce congestion itself in real-time. The HDRL is developed using Convolutional Neural Networks, and Recurrent Neural Networks combined with Proximal Policy Optimization (PPO) to be used as an adaptive traffic management system. The HDRL will take multi-modal traffic data (a combination of spatial vehicle density and temporal vehicle flows, etc.) into consideration, and when the active learning process has successfully trained the HDRL thoughtfully, it can make informed decisions based on previous experience to minimize average travel time (ATT), reduce congestion index (CI), reduce CO2 emissions, and improve safety. Our HDRL has validated its effectiveness against traditional fixed time signal control, rule-based adaptive systems (e.g., SCATS) and deep reinforcement learning (DRL) techniques, e.g., Deep Q-Networks (DQN), on the SUMO (Simulation of Urban MObility) platform modelled in grid, downtown, and highway locations respectively. This paper describes, in detail, the methodology applied and the results associated with it, providing representations of the results with six tables and eight graphics on the transformations of HDRL in relevancy to intelligent mobility solutions.