3. AgentsMovements

The generic vehicle movements in SimMobility ST simulator is based the MIcroscopic Traffic SIMulator (MITSIM) that was developed in the 90s for modeling traffic flows in networks involving advanced traffic control and route guidance systems at the Massachusetts Institute of Technology. The driving behavior model of MITSIM was then enhanced by several researchers to account for complex behaviors and different vehicle interactions. In SimMobility, the trips specified in a specific scenario are translated into individual vehicles wishing to enter the network at a specific time. Behavioral parameters (e.g., desired speed, impatience in following a slower vehicle, critical gaps in changing lanes, compliance rates to traffic signals and signs, etc.), information accessibility, and vehicle characteristic parameters are randomly assigned to each vehicle driver combination. SimMobility ST moves vehicles according to several driving behavior decisions. Driving behavior decisions are modeled as a series of interdependent choices that are based on a specific plan/tactic. For example, drivers select a target lane and adapt their acceleration and lane-changing actions to facilitate arriving at the chosen lane. The evolving circumstances (i.e., behavior of other drivers, traffic control) can cause changes to the plan. For example, drivers may initially plan to merge into mainline traffic through normal gap acceptance. But as they approach the end of the merging lane and are unable to find acceptable gaps, they may force merge. Drivers' plans are generally un-observable in the real-world (only drivers' actions are observed). The general framework of these models is shown in Figure below. At any instant, drivers choose a plan based on the state they are faced with. Their actions depend on the chosen plan. These actions, the actions of the other drivers and changes in the state of the control system (e.g., traffic light indications) may lead drivers to change their plans. The main driving behavior models in the plan-action structure are lane-changing and acceleration. In SimMobility, they are modeled using an integrated framework (see Section 6).

General decision structure

Each vehicle in SimMobility ST has its own destination and a path leading to that destination. Vehicles are moved at a fixed step size along their paths in accordance with various constraints. The information provided by the route guidance system acts on drivers' pre-trip and en-route route choice decisions through probabilistic route choice models. These models calculate the probabilities of choosing alternative routes at each intersection or decision point as a function of the perceived or provided information. Interactions with vehicles ahead, response to traffic controls, desired speed, and lane-use preference determine the movements of vehicles. These interactions are manifested in lane change decisions, acceleration and deceleration rates applied at any given time.