Planning to Chronicle


An overall view of a wildlife reserve, with 5 main regions, a grass field, a jungle, a command post, a riverside, and a lakeside. Notable fauna includes a cheetah, a crocodile, a herd of gazelles, and a flamboyance of flamingos; the latter two, being gregarious, remain as a group. Three robots, two drones and a ground vehicle, are tasked to cooperatively make a documentary from the events happen at the wildlife reserve. At each hour the animals move between the mentioned regions. The events of interest are gazelle grazing, cheetah eating gazelle, crocodile eating gazelle, flamingos mating, and crocodile eating flamingo. The robots does not know where the animal will go in the next hour, and thus, they should decide where to go for the next hour and which events try to capture to make a documentary whose sequence of events is specified by the user such that the expected time to make the documentary is minimized.

Description

This project focuses on the idea of deploying robots to observe a stochastic process in order to capture a sequence of events that meets specifications given by the user. This idea can be used in applications where we want to use robots to autonomously make a structured video or documentary from events that happen unpredictably in an environment. An example of those applications is where we want robots (rather than humans) to make a documentary from a wildlife reserve. The robot knows what kind of events may happen, but it does not know where and when those events will occur. It should decide which events to try to capture in the next time step. The chosen event should likely happen and it should be useful in making a useful story and the robot must be able to position itself to capture that event. The project proposes mathematical modeling for that purpose and solves in that context, several optimization problems, such as the problem of minimizing the expected time to record a desired video.

References

Hazhar's home page