1 / 81

Summary

“HUNTING CARIBOU HUNTERS BENEATH MODERN LAKE HURON 2 : USING VIRTUAL WORLD MODELING AND SERIOUS GAMES TO GUIDE THE EXPLORATION OF THE ALPENA-AMBERLEY LAND BRIDGE.”.

matteo
Télécharger la présentation

Summary

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. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. “HUNTING CARIBOU HUNTERS BENEATH MODERN LAKE HURON 2:USING VIRTUAL WORLD MODELING AND SERIOUS GAMES TO GUIDE THE EXPLORATION OF THE ALPENA-AMBERLEY LAND BRIDGE.” Robert G. ReynoldsProfessor Computer ScienceWayne State UniversityDetroit, Michigan 48202andAssociate Research Scientist Museum of AnthropologyUniversity of Michigan-Ann Arbor

  2. Summary • The Alpena-Amberley land bridge • Goals • Influence Maps • Path-finding • Flocking • Cultural Algorithms • Implementation • Results • Graphical User Interface • Future Work

  3. Early Human Occupation of the Upper Great Lakes • PaleoIndian (fluted point) occupation • Begins: ? • Ends: 10,300 BP with drop from Lake Algonquian (i.e. Lake Stanley) • Archaic 10,000-2,000 BP • Early and Middle Archaic poorly known (Lake Stanley) • Late Archaic sees major explosion in known sites (Post- Lake Stanley). • Tight correspondence assumed between PaleoIndian social organization, big game hunting, and the environment of the Lake Algonquian high water phase. • Very few known sites of Early and Middle Archaic Age.

  4. Lake Stanley Dates and Elevations

  5. Layered Search and Evaluation Strategy

  6. AUV and ROV

  7. Project Goals • Hunter behavior reflects that of their food sources, Caribou. • Knowing where Caribou are likely to be at given times of the year can be used to predict the location of human occupants. • Maintain a virtual world that can be used to visualize the past landscape that the surveyors are currently looking at.

  8. Extensive (Macro) • Can be a repository of found artifacts and structures stored in a “virtual” or cognitive database. • Produce a variety of “influence maps” that can contain information about present and past environments that are taken from running the “virtual world”. These can be used by surveyors to make “tactical decisions” about where to deploy the various survey devices.

  9. Intensive Survey (Meso) • Can be used to interpret “videos” taken by underwater devices in terms of the location of the device in the past virtual world. • GPS positioning of underwater device can be used to locate its position in the “virtual world”.

  10. Scale of Contribution (Micro) • Can be used to link up the view on the lake floor taken by a survey device with the corresponding portion of the landscape in the virtual world. • Virtual site prospecting. • Provide a virtual context for “found” artifacts and structures.

  11. Stages of Virtual World Development • 1)Pristine Environment-Caribou in Tundra environment. • 2)Add in human occupation. • 3)Adjust system to respond to long term climate change, e.g. lake levels shifts.

  12. Caribou Simulation Goals • Previous work focused on modeling tundra environment and individual animal behavior • Walters, ’75 – Barren ground caribou dynamics • Bergman ’00 – Caribou movement as correlated random walk • Bliss ’73, Price ’99, Sirois ’99 – Arctic tundra ecosystems • Other previous work by Reynolds and students in WSU AI lab began to take a look at the holistic view • Reynolds ’09 – Design of reality games • Vitale ’09 – Integration of environment interacting with caribou • More work ongoing

  13. Simulation Goals

  14. Simulation Goals • Use a “serious game” to tie together environment and entities • Game designed for use other than entertainment (Abt, ’70) • Extensibility • Reusability • Interactivity • Discover migratory paths and patterns • Where are the likely areas of local concentration • Can we predict behavior and actual historical sites

  15. Alpena-Amberley Land Bridge

  16. Simulation Goals • Utilize a number of technologies for effective simulation and presentation • Microsoft’s XNA Framework • Influence Maps • Path-Finding • Cultural Algorithms • Use multiple knowledge sources in tandem, which will be used and refined by Cultural Algorithms to generate realistic migratory behavior by herds.

  17. Game Program Overview

  18. Caribou Simulation Game • COMPONENT FRAMEWORK: • 3D virtual world component • Basic game engine component • CAT 3.0 component

  19. Pilot Study Search Areas

  20. Environment Generation • Combination of real-world and simulated data

  21. Environment Generation

  22. Environment Generation Data not always smooth

  23. Environment Generation • Simple method based on models of boreal growth • Two types of vegetation, tree and scrub. • Different levels of nutritional value. • Considered the availability of light and water. • More thorough modeling, such as seed transport, forest fires and age, may yield different maps

  24. Area 1

  25. Environment Generation

  26. Influence Maps – What They Are • The mechanism behind several knowledge sources • They are 2D or 3D cellular divisions of a world • Tactical values associated with each cell. • Strong support and history in gaming community (Tozour, ’01) • Thorough support system built around this module. • Dynamically alter and constrain values • Dynamic size and cell dimensions based on creation parameters • Load and save maps from and to greyscale • Merge and branch maps • Discover values at specified positions or indexes, etc

  27. Influence Maps – What They Are A 2D influence map Changing influences

  28. Influence Maps - Implementation Using multiple maps: by splitting maps, we can “save” a particular influence. Merging maps using negative values acts as deterrents.

  29. Influence Maps – Our Usage • We base our maps on certain parameters to use as input: • Availability of food. • Historical death rate. • Passability of terrain. • Initialized according to those parameters • Knowledge sources for Cultural Algorithm • Other parameters for future work: • Predator behavior • Wolves, humans, etc • Local influence • Drive lanes

  30. Basic AI Engine

  31. ALGORITHMS FOR STEERING SOFTBOTS IN GAME WORLDS • Pros • Can be searched using A★ heuristic search • Commonly used paths can be stored for quick access • Cons • Worst-case complexity is exponential in nature • Paths may look unrealistic and require post-processing in some cases GRID-BASED METHODS NAVIGATION MESHES REACTIVE METHODS AGENT-BASED APPROACHES HYBRID APPROACHES HARDCODED SYSTEMS

  32. ALGORITHMS FOR STEERING SOFTBOTS IN GAME WORLDS • Pros • Partitions the environment’s terrain into polygons • Waypoints are used to connect points that create paths • Cons • Works best with static environments • Similar issues to Grid-Based Methods GRID-BASED METHODS NAVIGATION MESHES REACTIVE METHODS AGENT-BASED APPROACHES HYBRID APPROACHES HARDCODED SYSTEMS

  33. ALGORITHMS FOR STEERING SOFTBOTS IN GAME WORLDS • Pros • Apply local environmental information to generate movement • Characters are react from intrinsic information within the landscape • Reactions are determined on a per-application basis, most often using “potential fields” • Cons • Bots often get stuck from poorly designed potential fields GRID-BASED METHODS NAVIGATION MESHES REACTIVE METHODS AGENT-BASED APPROACHES HYBRID APPROACHES HARDCODED SYSTEMS

  34. ALGORITHMS FOR STEERING SOFTBOTS IN GAME WORLDS • Pros • Apply local environmental information to generate movement • Characters are react from intrinsic information within the landscape • Reactions are determined on a per-application basis, most often using “potential fields” • Cons • Bots often get stuck from poorly designed potential fields GRID-BASED METHODS NAVIGATION MESHES REACTIVE METHODS AGENT-BASED APPROACHES HYBRID APPROACHES HARDCODED SYSTEMS

  35. ALGORITHMS FOR STEERING SOFTBOTS IN GAME WORLDS • Pros • Used often in computer graphics • Behavior is determined by: • a specified set of rules • social forces • particle swarm methods • Cons • Requires quantifying, identifying and controlling abstract knowledge and information GRID-BASED METHODS NAVIGATION MESHES REACTIVE METHODS AGENT-BASED APPROACHES HYBRID APPROACHES HARDCODED SYSTEMS

  36. ALGORITHMS FOR STEERING SOFTBOTS IN GAME WORLDS • Pros • Combines several approaches in one • Can consider local and global information • Cons • Difficult to design and implement • Requires in-depth knowledge of the problem GRID-BASED METHODS NAVIGATION MESHES REACTIVE METHODS AGENT-BASED APPROACHES HYBRID APPROACHES HARDCODED SYSTEMS

  37. ALGORITHMS FOR STEERING SOFTBOTS IN GAME WORLDS • Pros • Always engineered with the optimal path solution • Direct-control of agent steering • Cons • Very restrictive • Unresponsive to the smallest change • Non-autonomous GRID-BASED METHODS NAVIGATION MESHES REACTIVE METHODS AGENT-BASED APPROACHES HYBRID APPROACHES HARDCODED SYSTEMS

  38. Path Planning in Games with Cultural Algorithms: Previous Work • Super_Mario Competition (2010). • Car Racing Competition (2008).

  39. Mario AI Controller & Logistics

  40. Path Planning in games with ca • Reynolds & Kinnaird-Heether [2008] • WCCI 2008 Competition; • Socially motivated, agent-based approach: • Cultural Algorithms • 3D racing environment • Parameterizes rules for an single racecar driver • RESULTS: Steers a car around a track

  41. EXAMPLE OF CA OPTIMIZATION Reynolds & Kinnaird-Heether Applied CA to learn racing parameters Came in second at the WCCI 2008 How can this approach be be scaled up? acceleration(speedX, maxSpeed) if(speedX<maxSpeed) accel= 1 else accel= 0

  42. THE TALE OF TWO SIMULATIONS REYNOLDS/HEETHER RACECAR CARIBOU SIMULATION LOCATION: Alpena-Amberley Ridge CA INTERFACES WITH: Virtual World (1st party) GOAL: Multiple agent path-planning • LOCATION: • Multiple race tracks • CA INTERFACES WITH: • T.O.R.C.S (3rd party) • GOAL: • Single agent path-planning

More Related