Homepage Inria website

Section: New Results

Autonomous Virtual Humans

Space and Time Constrained Task Scheduling for Crowd Simulation

Participants : Carl-Johan Jorgensen, Fabrice Lamarche [contact] .

Crowd distribution in cities highly depends on how people schedule their daily activities. When performing an intended activity, people decisions and behavior mainly consist in scheduling tasks that compose this activity, planning paths between locations where these tasks should be performed, navigating along the planned paths and performing the scheduled tasks.

We proposed a task scheduling model aims at selecting where, when and in which order several tasks, representing an intended activity, should be performed. The proposed model handles spatial and temporal constraints relating to the environment and to the agent itself. Personal preferences, characterizing the agent, are also taken into account. Produced task schedules are optimized on the long term and exhibit adequate choices of locations and times with respect to the agent intended activity and its environment. Once computed, these task schedules are relaxed and used to drive a microscopic crowd simulation in which observable flows of pedestrians emerge from the scheduled individual activities. Such simulations are easy to produce and do not require the use of a complex decisional model. In terms of validation, we conducted an experiment that shows that our algorithm produces task schedules which are representative of humans' ones.

This work is part of the iSpace&Time project in which virtual cities are populated with virtual pedestrians and vehicles.

Long term planning and opportunism

Participants : Philippe Rannou, Fabrice Lamarche [contact] .

Autonomous virtual characters evolve in dynamic virtual environments in which changes may be unpredictable. One main problem when dealing with long term action planning in dynamic environment is that an agent should be able to behave properly and adapt its behavior to perceived changes while still fulfilling its goals.

We propose a system that combines long term action planning with failure anticipation and opportunism. The system is based on a modified version of an HTN planning algorithm. It generates plans enriched with information that enable a monitor to detect relevant changes of the environment. Once such changes are detected, a plan adaptation is triggered. Such adaptations include modifying the plan to react to a predicted failure and more importantly to exploit opportunities offered by the environment.

This system has been extended to better take into account the relationship between action planning and the environment. It is now combined with our space and time constrained tasks scheduling system (Cf. 6.5 ) to optimize the choice of locations where actions should be performed.