Presentation is loading. Please wait.

Presentation is loading. Please wait.

Agent Vision in 3D Environments Paul Werbicki Supervisor: Dr. Rob Kremer Department of Computer Science University of Calgary.

Similar presentations


Presentation on theme: "Agent Vision in 3D Environments Paul Werbicki Supervisor: Dr. Rob Kremer Department of Computer Science University of Calgary."— Presentation transcript:

1 Agent Vision in 3D Environments Paul Werbicki Supervisor: Dr. Rob Kremer Department of Computer Science University of Calgary

2 Agenda Background Proposal –Problem Statement –Statement of Work –Definitions –Why? –How? –Scope of Work Conclusion

3 Background – Possible Research Computer Graphics –Real-time rendering algorithms. –Environmental modeling and spatial data structures. Artificial Intelligence –Development of agents with human-like characteristics. –Interaction between agents and architectures.

4 Background – Computer Games Combine elements of algorithm optimization, data structures, graphics rendering, networking, and artificial intelligence. Closely simulate human-like environments. Gaining wider-spread use as a research tool (Manojlovich, Chen and Lewis, 2003; Laird 2002).

5 Proposal – Problem Statement How do agents simulate vision in a virtual 3D environment? –Ray casting –Waypoints –Network paths –Area Awareness System (J.M.P. van Wareren, 2001)

6 Proposal – Vision Example Puzzle.avi – Area Awareness System (J.M.P. van Wareren, 2001)

7 Proposal – Current Limitations Analogy: exploring a dark room. Pre-calculated spatial information. Combination of visual information with cognitive model. Pure separation of spatial information from environment navigation difficult.

8 Proposal – Statement of Work I propose using Computer Graphic rendering techniques, such as Binary Spatial Partitioning (BSP) Trees, to render a 3D environment into a data structure usable by an agent, thereby allowing it to “ see ” that environment.

9 Proposal - Definitions Agent – is anything able to perceive and act upon its environment. Vision – is time and position dependent sensory information, based on an agent’s field of view in an environment. Agent Architecture – is the system that manages the interaction between the environment (vision) and an agent.

10 Proposal - Why? Real-time generation of visual information. Level playing field between human agents and software agents. Simulate human-level perception of an environment. Provide agent with pure spatial information leaving environment navigation up to implementer.

11 Proposal - How? Four stage approach: 1.Model environment, group polygons and provide meta information (stairs, doorway, etc.) 2.Pre-compute BSP Tree of the environment. 3.Transverse the BSP Tree based on field of view and send polygon and meta information to the agent. 4.Allow the agent to interpret the information and effect the environment (move, pick up object, etc.)

12 Proposal - Scope of Work Goal is to study agent architectures and the use of real-time graphic rendering techniques in delivering visual information. Implementation will cover first three stages. Fourth stage left as possible Ph.D. research topic.

13 Conclusion Fun and interesting work that crosses Computer Graphics, Artificial Intelligence and video games. Framework for future research into spatial perception, memory, motion planning, and agent learning. Help converge innovations between the game industry and the research community (Laird, 2000).

14 References J.M.P. van Waveren. The Quake III Arena Bot 2001. University of Technology Delft. John E. Laird, et al. “ A Test Bed for Developing Intelligent Synthetic Characters. ” AAAI Spring Symposium on AI and Interactive Entertainment (Mar 25-27, 2002). John E. Laird. “ Research in Human-Level AI using Computer Games. ” Communications of the ACM 45.1 (Jan 2002): 32-35. Accessed on 23 Sep 2002.http://doi.acm.org/10.1145/502269.502290 John E. Laird and Michael Van Lent. “ Human-Level AI ’ s Killer Application: Interactive Computer Games. ” Proceedings of the Seventeenth National Conference on Artificial Intelligence (2000): 1171- 1178. Accessed on 1 Oct 2003 http://www.kddresearch.org/Groups/AI-CBR/Papers/Interactive- computer-games-AIMag22-02-003.pdf John E. Laird. “ Bridging the Gap Between Developers and Researchers. ” Game Developer (Aug 2000). Accessed on Oct 21 2003 http://www.gamasutra.com/features/20001108/laird_03.htm Joseph Manojlovich, Jinlin Chen and Michael Lewis. UUTSAF: A Simulation Bridge between OneSAF and the Unreal Game Engine 2003. Accessed on Oct 30 2003 http://usl.sis.pitt.edu/ulab/pubs/utsaf- smc03.pdf Id Software Inc. Quake II 1997. Accessed on Oct 18 2003 http://www.idsoftware.com


Download ppt "Agent Vision in 3D Environments Paul Werbicki Supervisor: Dr. Rob Kremer Department of Computer Science University of Calgary."

Similar presentations


Ads by Google