snowpuppy / augreality Goto Github PK

Ignore. Accidental duplicate.

So far we have planned on storing real-world obstacles in the form of lat/long coordinates of corners. However, now I wonder if it would make more sense to let the CCU use the binary image directly, which presents the following tradeoff:

More memory

up to 50MB without encoding. possibly under 1MB with run-length encoding.

Less runtime computation

Based on the two scenarios I can forsee us needing obstacle information:

• Verifying that players and newly-generated game objects (e.g. Pacman fruit) are 'in-bounds'. With images, this is as simple as isPixelObstacle(LatToX(lat),LngToY(lng)). But with obstacle corners, this will require checking whether the lat/lng coordinate is on a certain side of all lines generated from the corners.
• Updating game object locations (i.e. game AI). In this scenario, I am not sure which obstacle format is more useful. Thor, I assume you will play a large part in the Pacman ghost AI: would it be easier to work with pixels or obstacle corners for game AI?

The following are conservatively large estimates of the necessary area needed for a game area, at two different zoom levels. 633KB (z=19) and 1282KB (z=21), respectively.