directable multi agent navigation and simulation l.
Download
Skip this Video
Loading SlideShow in 5 Seconds..
Directable Multi-Agent Navigation and Simulation PowerPoint Presentation
Download Presentation
Directable Multi-Agent Navigation and Simulation

Loading in 2 Seconds...

play fullscreen
1 / 8

Directable Multi-Agent Navigation and Simulation - PowerPoint PPT Presentation


  • 454 Views
  • Uploaded on

Directable Multi-Agent Navigation and Simulation Sachin Patil Project Proposal Authoring very large crowds is challenging Simulating emergent behaviors using agent-based methods is non-trivial Setting up simulations to get desired output is tedious

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about 'Directable Multi-Agent Navigation and Simulation' - Ava


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
project proposal
Project Proposal
    • Authoring very large crowds is challenging
    • Simulating emergent behaviors using agent-based methods is non-trivial
    • Setting up simulations to get desired output is tedious
  • Need to develop real-time tools for authoring multi-agent simulations
    • Degree of control can vary from scripting individual behaviors to controlling crowd flows (macroscopic)
problem definition
Problem Definition
  • User-guided navigation is a step towards directable crowds
    • Top-down approach
    • Provide macroscopic control over navigation of groups of agents using vector fields (flow fields)
    • Specification of emergent behaviors such as formation of lanes and vortices
    • Intuitive UI for editing multi-agent simulations
previous work
Previous Work
  • Authoring crowd simulations
    • Steering Behaviors for Autonomous Characters [Reynolds 99]
    • CrowdBrush : Interactive Authoring of Real-Time Crowd Scenes [Ulciny 04]
    • Scalable Behaviors for Crowd Simulation [Sung 04]
    • Flow Tiles [Chenney 05]
    • Interactive Control of Large Crowd Navigation in Virtual Environments using Vector Fields [Jin 08]
    • High-Level Path Specification and Group Control [Silveira 08]
    • Group Motion Editing [Kwon 08]
    • Massive [MassiveSoftware] has utilities to paint flow fields on terrain to direct crowd flow (http://www.youtube.com/watch?v=e0SIezN-0UU)
approach
Approach
  • Design vector fields for user-guided navigation
    • Construct local navigation field for every group of agents (goal-driven)
    • Global vector field specification for the entire scene (lanes, vortices and other movement patterns)
  • Use a grid-based approach
    • Solve the eikonal equation on a regular grid using FMM
    • Modify cost function based on user input
    • Parallelization over many cores/GPU
      • A Fast Eikonal Equation Solver for Parallel Systems, Jeong and Whitaker, SIAM-SC 2007
  • Flow fields for global navigation + RVO for local collision avoidance!
    • Overall motions not smooth enough (modify RVO penalty criteria)
novel aspects
Novel Aspects
  • Designing arbitrary vector fields for authoring multi-agent simulations
  • Novel cost function for the eikonal equation
  • Eikonal equation needs to be solved per user-edit and not per simulation step
  • Intuitive interface for authoring simulations (user mouse gestures, paint-brush like interface etc.)
project roadmap
Project Roadmap
  • Incorporate new cost function in FMM based solver
  • Develop intuitive UI for editing simulations
  • Speedups
    • Possibly port to many-cores/GPU based on [Jeong 07] for interactive editing
    • E* to locally repair fields instead of re-computing the entire solution
  • Extensions:
    • Tool for higher-level behavior specification (queuing, following, shepherding etc.)
    • Dynamic flow fields
    • Data-driven crowds