- In 2017, the cost of congestion in the United States was around 305 billion dollars, and city-dwellers, on average, lost 1400 dollars while sitting 42 hours in traffic jams. Aiming for better mobility and more efficient utilization of the transportation network, emerging connected and autonomous vehicle (CAV) technologies and their communication capabilities can produce well-coordinated and more efficient routing behavior to dissipate the traffic rather uniformly throughout the network, resulting in slower travel times. Vehicle routing is among the most critical and challenging, yet unsolved, tasks in CAV research. Current routing strategies either rely on a centralized control system which can fail in scaling, or employ decentralized schemes that yield sub-optimal coordination and poor system performance. In addition, it is of great importance for the deployment of CAV technologies to understand the transportation systems behavior in a mixed environment with various levels of communication complexity, where CAVs and Non-CAVs coexist and interoperate. The routing problem in a multiagent system resembles a competitive congestion game. The decisions of one agent (in this case, a CAV) directly impacts the performance of the others. When the number of agents traversing the same transportation facility at the same time exceeds a certain threshold, bottlenecks may occur, and thus, higher travel times. Therefore, coordination between CAVs is key to avoiding such circumstances. This dissertation answers how and to what extent different routing optimization algorithms, under various levels of autonomy and communication capabilities, can increase the mobility of the transportation system. This work designs this system in a decentralized manner that scales linearly in achieving a social and system-level optimum. To realistically analyze this system, we investigated the coordination behavior of CAVs under (1) No Communication, (2) Minimal Communication, and (3) Extensive Communication. In the absence of connectivity between the CAVs, a learning-based approach has been implemented where each CAV optimizes its own route using a reinforcement learning technique and based on its prior experiences. This competitive game quickly overwhelms the system as the market penetration of CAVs surpasses the critical threshold range (50% to 75%), where the mobility improvements are the most significant, and beyond which the system performance degrades. Under minimal communication level, we assumed the CAVs share information regarding their location and speed with the rest of the CAVs in their communication cluster through a multi-hop network. Then, a coordination scheme was implemented where each CAV minimizes its travel time based on the limited information it receives. Results showed that this application can reduce system travel time by up to 20%. Additionally, the emergence of mobility benefits are shown to correlate with the CAV network characteristics through the lens of percolation theory. The results revealed that, for the mobility benefits to surface, at least 70% of the CAVs are required to form a communication cluster. Under an extensive communication capability, where the CAVs not only share their location and speed but also their preferred path to their destination, a reduction of up to 40% in system travel time was achieved for high levels of CAV market penetration and communication radius. Moreover, the improvement in mobility was proved to be highly associated with the uniform dissipation of traffic onto the network. These findings provide solid support to create evidence-driven frameworks to guide future CAV development and deployment in a decentralized and coordinated manner.