Urban traffic is increasingly diverse and unpredictable. Cars, buses, motorcycles, bicycles, and (e-)scooters interact within limited road space, creating complex collective dynamics that classical traffic models fail to capture. To study these phenomena, we analyze the pNEUMA dataset, where drones recorded detailed vehicle trajectories. The raw data contain noise from occlusions, perspective distortions, and annotation errors, so we apply a denoising and anomaly detection framework that preserves genuine maneuvers and rare events. After reconstruction, trajectories can be map-matched reliably, allowing extraction of lane-changing and other maneuvers. Analysis reveals mode-dependent lane discipline, with motorcycles forming self-organized lanes between cars. Longitudinal displacements during maneuvers follow a consistent pattern, and we also find that maneuverability decreases with speed, providing a key modeling ingredient. Maneuver durations are shown to follow a lognormal distribution, and a generalized time-to-collision metric shows heavy-tailed interactions.

Building on these observations, we introduce novel traffic flow models to link maneuverability with the speed differences between vehicle types and to simulate the collective behavior of multispecies traffic. These simulations reveal a transition from ordered lane formation to disordered clusters and identify bivariate scaling laws, providing evidence that multispecies traffic dynamics belong to a universal class of nonequilibrium systems. Finally, we introduce an analytically tractable lane-changing model in which maneuver duration emerges from maneuverability and safety constraints. Lane changes are treated as discrete decisions with continuous execution times and can be interpreted as a generalized intersection problem, where post-encroachment time ensures collision-free maneuvers. Importantly, this model endogenously reproduces the empirically observed lognormal distribution of durations, thus connecting microscopic behavior to macroscopic traffic phenomena.