Observing to infer

Supplementary movies for the paper titled,

'Observing a group to infer the individual characteristics'

Arshed Nabeel and Danny Raj M

 Movie S1. Clustering, Lane formation and Jamming in bi-disperse crowds

Bi-disperse crowds consist of two groups of agents, where each agent has an intrinsic direction of motion. Interactions with other agents (which depends on the number of agents in each group and their speeds) results in a wide variety of collective phenomena from clustering (where the agents get agglomerated to an absorbing state), laning (where agents of each group takes up a distinct lane(s)) to jamming (where the movement is no longer counter-current).


Movie S2. As the intrinsic speed increases, minority agents (hitherto trapped) begin to break free.

Agents in the minority group get trapped when their intrinsic speed is very small. The entire assembly is jammed and drifts with a velocity which corresponds to the mixture velocity of the frozen assembly. As the intrinsic speed increases, these trapped agents begin to break free and in the process slow down the agents in the majority group to give rise to an effect similar to the so called, faster-is-slower effect.

Movie S3: Working of a neighbourhood classifier and its failure mode.

A neighborhood observer classifies the agent as belonging to one group or another using (i) the velocity of agent and  (ii) a network parameter that characterizes the role of the neighborhood on the motion of the agent. The neighborhood classifier is able to differentiate the agents in the two groups even when they move in the same direction. The neighborhood classifier performs poorly when the two groups are symmetric.