Increased response time of emergency vehicles (EVs) can cause an irreparable loss of life and property. Reducing emergency services’ response times requires identifying the most effective optimization and pre-emption techniques and factors for reducing EVs travel times. Research in this domain has adopted one or both of optimization and pre-emption techniques for routing EVs. None of the existing studies provides any comparative evidence that a particular optimization technique and pre-emption system is better suited to reducing the EV’s travelling time.

An appropriate solution to improve existing techniques requires dynamic optimization and efficient and precise pre-emption to cause minimal disruption to other vehicles. The success of such dynamic optimization and pre-emption systems depends on the availability of real-time dynamic traffic data. This means that sensors deployed at various road network infrastructures must communicate in real-time and support real-time decision-making. In general, traffic infrastructure requires a deeper integration with software systems to ensure high availability of accurate real-time data. This thesis’s main focus is to develop precise pre-emption and dynamic optimization techniques that can aid in reducing the response time of EVs. The main contributions of this thesis are: • a conceptual model to analogically map traffic scheduling with scheduling algorithms in real-time systems. • an adaptive fair scheduling algorithm (FSA) that ascertains higher travel time reliability developed by analogically mapping traffic domain with real-time systems. • an emergency vehicle pre-emption (EVP) technique with different levels of priorities that ascertains a certain level of performance assurance to different criticality levels in emergency services. • a decentralized self coordinating traffic system to prioritize emergency vehicle movement through an isolated traffic intersection using Virtual traffic lights plus for emergency vehicles (VTL+EV) algorithm. • an ensemble-based model that learns from multiple pre-engineered features related to topology, directions, the shape of the trajectory, path, speed limits, map distance traversed, real-time traffic conditions and precisely estimate the travel time.

We conducted exhaustive experimentation, and empirically proved that the FSA is more reliable and fairer in scheduling in terms of travel reliability compared to existing state of-art algorithms in terms of buffer index and Jain’s fairness index. Our experimental results confirm that the EVP algorithm can significantly reduce the average waiting time of regular traffic and ensure all EVs meet their target response time. Comprehensive experiments and results showed that VTL+EV has the evident advantage of reduced waiting time for regular traffic and emergency vehicles in both congested and non congested traffic conditions. We also concluded that for wisely extracted manual features, ensemble-based gradient boost regression approach could outperform existing state-of-art baseline models that employ deep neural networks in travel time estimation.