Pedestrian Motion

Pedestrian positioning and motion is modelled in the 2-dimensional Cartesian frame using simulation coordinates. Since pedestrian motion is omnidirectional and not strictly bound by road lanes, there is no need to translate their motion to a different frame, such as the Frénet frame which is used for vehicles.

During simulation, the dynamic state of pedestrians is represented by a vector containing their current position, speed, acceleration, and heading angle. This vector is directly updated by the environment at each simulation cycle. The point-mass model used for pedestrians abstracts certain characteristics such as demographic information and pose. This representation simplifies the movement calculations done by the Social Force model.

The Social Force model used in the Motion Planner layer of the SP model incorporates four main forces: an attracting force towards the pedestrian's goal point (f_adapt), repulsive forces from vehicles (f_ik) and other pedestrians (f_ij), and repulsive forces from the walls or borders in the environment (f_iW). These forces are summed to generate a resulting force acting on the pedestrian. The figure below shows a high-level visualization of these forces.

In a traffic environment, pedestrians cannot necessarily walk in a straight line directly to their goal point. SP pedestrians dynamically select intermediate waypoints that they must visit in order to navigate through the environment. Some examples of common waypoints include entrances and exits of crosswalks and points along a section of sidewalk. At a given time, each SP pedestrian has one waypoint that is used by the Social Force model calculation as the goal point.

The model contains a local planner, in the Motion Planner layer, with which pedestrians plan a local path from their current position to their waypoint. The local path consists of an ordered sequence of connected lanelets (sharing endpoints). Each time the pedestrian reaches their current waypoint (the end of the local path), a new waypoint is selected and a new local path is generated. This process emphasizes the importance of running simulation scenarios on highly accurate Lanelet2 maps containing all the necessary pedestrian-specific lanelets outlined in the Pedestrians section.