Stay off the lawn - Creating smooth paths based on region - - PowerPoint PPT Presentation

Stay off the lawn - Creating smooth paths based on region preferences

Norman Jaklin1 Atlas Cook IV2 Roland Geraerts1

1Utrecht University - Department of Information and Computing Sciences 2University of Texas at Austin - Institute for Computational Engineering and Sciences

ASCI.OPEN / ICT.OPEN 2013

Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 1 / 19

Overview of our research Path planning and crowd simulation for virtual autonomous agents

General framework based on Explicit Corridor Map [1]

Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 2 / 19

Overview of our research

Research on the data structure itself

Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 3 / 19

Overview of our research

Research on virtual crowds

Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 4 / 19

Overview of our research

Combining individual steering behavior with coordinated crowd movement

Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 5 / 19

Motivation

Problems with terrain-based path planning in current games and simulations

World of Warcraft by Blizzard Entertainment 2004 Grand Theft Auto IV by Rockstar North 2008 Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 6 / 19

Problems with terrain-based path planning in current games and simulations

Inflexible solutions for handling individual terrain preferences

Traversable areas treated as equal, independent of individual terrain preferences Unattractive traversable terrain is made impassable for particular characters

Grand Theft Auto IV by Rockstar North 2008 Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 7 / 19

More problems

Paths are often illogical and not visually convincing

Unnecessary detours No smooth trajectories Unnatural clearance from obstacles Characters do not look ahead enough Obstacles or rough terrain is ignored

World of Warcraft by Blizzard Entertainment 2004 Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 8 / 19

MIRAN - Modified Indicative Routes and Navigation

Input:

A 2D polygonal environment Virtual characters with sets of individual region preferences An indicative route that roughly guides a character

automatically computed or manually designed

Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 9 / 19

MIRAN - Modified Indicative Routes and Navigation

Output: A natural-looking path that

gives the user control over the amount of smoothing is based on a character’s region preferences keeps clearance from obstacles avoids unnecessary detours can be computed in real-time

Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 10 / 19

The MIRAN method - Overview

Step 1: Compute reference point on the indicative route Step 2: Compute set of candidate attraction points Step 3: Pick best attraction point Step 4: Move character towards attraction point

Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 11 / 19

Two user-controlled parameters

The shortcut parameter σ The sampling distance d

σ d

character position indicative route

Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 12 / 19

Step 1: Compute reference point r

ri := first closest point on the indicative route between former reference point ri−1 and former attraction point αi−1

xi ri indicative route ri−1 αi−1

Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 13 / 19

Step 2: Compute set A of candidate attraction points

Visible points along the indicative route between ri and σi discretized with sampling distance d

xi ri σi

αi1 αi2 αi3 αi4 αi5 αi6 αi7 start goal

Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 14 / 19

Step 3: Pick best attraction point α from A

Each line segment between x and αj is weighted with the underlying type of terrain and the curve length distance from r to αj. Lower terrain costs ⇒ lower weight αj further ahead on the indicative route ⇒ lower weight

r x α1 α2 α3

start goal

Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 15 / 19

Step 4: Move character towards α

x1 x2 x3 x4 α1 α2 α3

Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 16 / 19

Forest example

Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 17 / 19

Future work

Improve the computation of indicative routes Handle disc-shaped characters with variable radius Extend terrain-based planning to local collision-avoidance routines Use continuous set of candidate points instead of sampling the indicative route Generalize MIRAN to (multi-layered) 3D environments with height information

Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 18 / 19

Thank you!

• R. Geraerts.

Planning short paths with clearance using Explicit Corridors. In Proceedings of the IEEE International Conference on Robotics and Automation, pages 1997–2004, 2010.

• N. Jaklin, A. Cook IV, and R. Geraerts.

Real-time path planning in heterogeneous environments. Computer Animation and Virtual Worlds (CAVW), 24:285–295, 2013. Jaklin et al. (Utrecht University) ASCI.OPEN / ICT.OPEN 2013 19 / 19