Multi agent intersection management considering energy consumption

Traffic congestion is one of the main reasons of increasing pollution and fuel consumption in the cities. According to recent urban mobility reports, traffic congestions cost up to $121 billion in USA every year due to productivity and time loss. Approximately 38-40 hours are spent by an average citizen in the traffic. Additionally, 25 billion kilograms of carbon dioxide is emitted due to congestions. Increasing number of vehicles on the urban roads leads congestions at the intersections. Traditional intersection management methods such as stop signs, traffic officer control and traffic lights, are getting insufficient. Traffic light timing depending on day time is a common application; however, it is not an adaptive method for changing traffic flow density. Traffic density can be measured by cameras, piezo and infrared sensors. Intelligent traffic light timing can be achieved using the traffic information gathered from the units integrated to the roads. Vehicle to vehicle and vehicle to infrastructure communication technologies allow us to gather information directly from vehicles, and analyze traffic congestion. Besides, vehicles can be informed about the traffic ahead or can be conditioned to reduce effect of congestions. Intersections can be managed by using information from the vehicles. One way to manage intersections is to integration of an intersection manager unit. Vehicles request reservation from intersection manager to pass the intersection. Another approach is interactive multi agent intersection management by the help of autonomous vehicles. Autonomous intersection management uses estimated trajectories provided by the vehicles to detect any possible crash at the intersection. Estimated position of a vehicle at a time can easily be simulated for a given velocity profile. Estimated position and time information of vehicles are compared and estimated collisions are detected. Possible collisions must be resolved before the vehicles arrive at the intersection. Most common approach is the first come – first served method, which allows the first vehicle that requests reservation first to pass the intersection, allocate the intersection. Other vehicles' trajectories must be adjusted to avoid the collision at the intersection. This adjustment is done by assigning delay to the arrival time of the vehicles at the intersection. Giving the priority to the vehicle that requests reservation first may not always be the most efficient decision in terms of total delay time. A method called look ahead intersection control policy is explained in the next sections. Look ahead intersection management policy assigns the priorities based on total delay time minimization. This method aims to reduce consecutive effect when the head vehicle of a convoy is delayed. In this study, we proposed an intersection management method searching the passing sequence from the collision point to minimize defined cost functions. First intersection model is explained to specify communication zone, velocity adjustment zone and grid structure at the center of intersection. Communication protocols are given for the communication zone. Crash detection and intersection management algorithm are executed in communication zone by all vehicles. Vehicles' estimated trajectories are transformed into time – space occupation at the intersection using grids to easily isolate possible collisions. Passing sequences of the vehicle, which are estimated to pass the collision point, are found. For each sequence, different vehicles' arrival are delayed by different amount of time. Therefore, each sequence results in different total delay time. Selection between possible passing sequences is done by minimizing total delay time. Another performance criterion is energy loss of the vehicle. Vehicle longitudinal dynamics are explained for energy loss calculation. Delay time is realized by deceleration and acceleration. Acceleration requires an increase in traction force, this leads energy loss compared to the nominal state. Since vehicle dynamics are different for different vehicles, the same velocity rate may result in different energy losses. Hence, each passing sequence has a different energy loss value associated with it. This difference is used to select the sequence with minimum energy loss. Energy loss and total delay based costs are then combined in one cost function using a rating parameter. After specifying cost functions, intersection management is simulated for different cases. A simulation framework is created in MATLAB. Vehicle kinematic bicycle model, which is used in the simulation environment, is explained. Vehicles with different masses and paths are simulated in different case studies. The case studies show that total delay based sequence selection distributes minimum delays to the vehicles only aiming to resolve estimated crashes. On the other hand, energy loss based method gives the priority to heavier vehicles to minimize energy loss. Thus, the total delay time increases in energy loss based method. The combined cost based selection method reduces energy loss of the total delay time based method. Similarly, it reduces total delay time of the energy loss based method. In this thesis, literature survey and motivation of the study is given. Then vehicle models are explained for further use in the thesis. The intersection model including communication protocols and crash detection are stated. Afterwards, the intersection management algorithm is explained. The total delay time, energy loss and combined cost functions are given. Finally, simulation environment is explained and results of case studies are discussed.