The goal of this article is to present a new approach to providing support to support-decision sharing in transportation systems through the development of Adaptive Sparse Multi-Objective RL (MORL) as a support for the traditional policies that have often been used in controlling traffic flow. By creating an Adaptive Sparse-MORL policy, we can create the ability to have a policy that can have the ability to respond to changes in real-time while still being able to maintain the characteristics of a policy. This allows us to enhance our current understanding of how to create an optimal policy in managing emergency vehicles as well as improving efficiency when providing timely assistance. Additionally, by providing additional tools by developing a Traffic Aware Policy Pruning Module which can adaptively reduce the size and complexity of the models used; this, in turn, aids in the reduction of the total number of resources used and improves the response time of our models. The use of a Pareto-Optimal Envelope Mechanism allows for the re-evaluation of the decisions made based on defined trade-offs between return time and network flow, thus providing a significant reduction of the computational workload. Furthermore, our Distributed Sensor Fusion Layer provides a means to merge heterogeneous sensor data through the use of cross-attention, resulting in a higher accuracy of detection without the reliance on any heuristic localization. The deployment of the system on the NVIDIA Jetson AGX and Orin edge devices is designed to enable interoperability with the existing traffic control systems via an SAE J2735 compliant Signal Phase and Timing Message. In addition, we incorporate a Fail-Safe Operation through local storage of the policies on the Jetsons in the event of a failure of the cloud-edge partitioning. The evaluation of our method indicates a reduction in the amount of unnecessary motor computations of 62% over the Fixed Interval MORL method while achieving a service availability rating of 98.7% during the degradation of the network. The implementation of TinyML for lower priority processing tasks will ensure continued performance even in cases of high-priority events, thus making it suitable for deployment in Intelligent Transportation Systems.