Factors that contribute to collision avoidance behaviours involving a single pedestrian versus a group of pedestrians in a natural environment.

Along busy walking paths many pedestrians walk in a group, thus challenging the collision avoidance behaviours of both the group and individual pedestrians. Here we sought to determine what factors contribute to these behaviours in a real-world environment. We hypothesized that pedestrian group size, spatial separation between parties, presence of an additional interferer not involved in the interaction, pedestrian distractedness, constraints to a pedestrian's mobility (i.e., holding or pushing an object), and pedestrian age would significantly influence the likelihood and extent of path deviations during pedestrian interactions. We analyzed videos of unscripted pedestrian walking behaviours on a busy urban path where a single pedestrian approached and crossed paths with a group. We applied deep learning algorithms to detect and extract pedestrian walking trajectories, had unbiased raters characterize pedestrian interactions, and used a multiple regression analysis to determine factors that influenced behaviours. Our analysis revealed that a smaller medial-lateral separation between approaching pedestrians and lack of an additional interferer in the area predicted a higher likelihood of path deviation for one or both parties. Furthermore, the presence of an additional interferer in the area, smaller group size, and if pedestrians were distracted were associated with greater medial-lateral separation at time of crossing. Overall, these findings may offer valuable insights for refining computational models of collision avoidance and improving the control algorithms of navigating robots that interact with human pedestrians.