makarov ilya, tokmakov mikhail, tokmakova lada - imitation of human behavior in 3d-shooter game

Post on 27-Jul-2015

324 Views

Category:

Presentations & Public Speaking

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Imita&on  of    Human  Behavior    

in  3D-­‐Shooter  Game    

Makarov  Ilya  Tokmakov  Mikhail  Tokmakova  Lada    

Can't  Kill  Progress  

Фон  

•  Brief  Review  of  Exis<ng          3D-­‐shooters    •  Geometry  •  Visual  Recogni<on  •  Decision  Making  Model  •  Process  of  Gaming  •  Conclusion  

AIST'2015,  Yekaterinburg   2  

Фон  

Brief  Review  of  Exis<ng  3D-­‐shooters    

AIST'2015,  Yekaterinburg   3  

Wolfenstein   Call  of  duty:  Ghosts  

Idea  

AIST'2015,  Yekaterinburg   4  

The  main  goal  of  our  project  is  to  replace  commonly  used  automa<c  target  recogni<on  and  target  aiming  by  using  visual  recogni<on  methods  and  to  test  obtained  model  

We  create  an  automa<c  decision  making  module  to  search  for  enemies  in  a  3-­‐D  maze  

Geometry    

•  Types  of  the  objects:    – walls,  boxes  –   columns,  doorways  

•  Processing  the  queue:  –  finding  element  with  maximal  priority  

–  recalcula<ng            dangerous  zones  –  comparing  the  first  K            queue  elements  

AIST'2015,  Yekaterinburg   5  

Example  

AIST'2015,  Yekaterinburg   6  

The  density  func<ons  represent  the  process  of  visual  recogni<on  of  objects  when  we  search  for  an  enemy  but  have  to  spend  some  <me  on  processing  object’s  shapes.  

Visual  Recogni<on    Screenshot with filled holes

AIST'2015,  Yekaterinburg   7  

Filtering of the screenshot

Yflgbcmyflj  pfrhsnm   Yflgbcmyflj  pfrhsnm   Yflgbcmyflj  pfrhsnm  

Yflgbcmyflj  pfrhsnm  

Decision  Making  Model  1)  Walking  

a)  Bypassing  and  Search  b)  Moving  toward  some  point  in  the  maze  c)  Finding  cover  

2)  Shoo<ng  a)  Sigh<ng  b)  Opening  fire  c)  Correc<on  rela<ve  to  recoil                        and  motor  reflexes    

3)  Visual  recogni<on  a)  Recogni<on  of  objects  b)  Recogni<on  of  dangerous  zones  c)  Recogni<on  of  enemy    

(shoo<ng  aeer  enemy  recogni<on)    

AIST'2015,  Yekaterinburg   8  

Process  of  Gaming    •  The  rela<ve  error  depends  on:  –  the  angle  between  BOT’s  rota<on  movements  and  direc<on  on  the  target  

–  X-­‐speed  –  Y-­‐speed    

AIST'2015,  Yekaterinburg   9  

Conclusion  

In  the  current  state  we  evaluate  the  parameters  to  iden<fy  the  dangerous  zones  and  to  sight  on  enemy.  We  proceed  with  the  comparison  of  the  methods  of  visual  recogni<on  to  improve  our  model.    

AIST'2015,  Yekaterinburg   10  

Thank  you  for  aien<on!    

Department  of  Data  Analysis  and  Ar<ficial  Intelligence  Na<onal  Research  University  Higher  School  of  Economics  

 Makarov  Ilya:  iamakarov@hse.ru  

   Tokmakov  Mikhail:  matokmakov@gmail.com        Tokmakova  Lada:  lrtokmakova@yandex.ru  

   

©  hip://www.square-­‐enix.com    

         

AIST'2015,  Yekaterinburg   11  

top related