What pathfinding approaches for an RTS?

I'm trying to add better pathfinding for the 3 ground units per side in a real-time strategy game for NES. The units move in a 28 by 24 cell map and cannot pass through solid cells, many of which are defensive walls built in the base building phase and destroyed piece by piece in the combat phase. One player's base is on one side of the map and the other on the other side. Currently each unit seeks a target location in the map by alternating horizontal and vertical steps toward the target, and if both are solid, it just gives up. Guiding the units around walls and other obstacles requires quite a bit of tedious micromanagement by the player.

About 704 of the 2048 bytes of work RAM are spoken for before I even start making a particular game.

• 16 bytes: local variables
• 176 bytes: sound driver and basic machine state
• 192 bytes: VRAM transfer buffer
• 64 bytes: CPU stack
• 256 bytes: shadow OAM

This leaves 1344 ($540) bytes. I can give a more detailed breakdown of game state if it'd help.

• 768 bytes: map, 32 bytes per row by 24 rows
• 434 bytes: game state, with base building and combat phases using overlapping allocations
• 142 bytes: free

What kind of data structures and algorithms could I use to step toward a target location more intelligently? Would it help to cache something and somehow update the cache every time a defensive wall is breached?

Statistics: Posted by tepples — Sat Apr 06, 2024 7:01 pm — Replies 2 — Views 114

---