Sliding window, replay tape, and normalization #83

New Issue

2026-08-25T19:11:37+02:00

SirStone commented

2026-08-25 19:11:37 +02:00

Parent

What to build

Data collection layer that runs inside the bot's tick loop. Each tick: record enemy state, normalize it, maintain two buffers.

Sliding window: last 30 ticks of enemy state (lateral_vel, heading_delta, wall_distance) as a flat 90-element buffer + current distance as 91st input. This is the ANN's input vector.

Replay tape: rolling ~2000-tick buffer of full enemy states for offline GA fitness evaluation.

Normalization: per-feature fixed scaling using game constants — lateral_vel / 8.0, heading_delta / π, wall_distance / 1000.0. No runtime statistics.

Acceptance criteria

Sliding window fills with normalized enemy state each tick
Window produces a flat 91-element float array ready for ANN input
Replay tape records raw enemy states up to ~2000 ticks, oldest dropped when full
Normalization uses game constants only — no running stats, no allocations per tick
Self-check: when isMainModule with doAssert validates normalization ranges and buffer behavior
Integrated into bot skeleton's tick loop — data collection runs every tick

Blocked by

#81

## Parent https://git.fossellini.top/SirStone/SirRoboGarage/issues/80 ## What to build Data collection layer that runs inside the bot's tick loop. Each tick: record enemy state, normalize it, maintain two buffers. Sliding window: last 30 ticks of enemy state (lateral_vel, heading_delta, wall_distance) as a flat 90-element buffer + current distance as 91st input. This is the ANN's input vector. Replay tape: rolling ~2000-tick buffer of full enemy states for offline GA fitness evaluation. Normalization: per-feature fixed scaling using game constants — lateral_vel / 8.0, heading_delta / π, wall_distance / 1000.0. No runtime statistics. ## Acceptance criteria - [ ] Sliding window fills with normalized enemy state each tick - [ ] Window produces a flat 91-element float array ready for ANN input - [ ] Replay tape records raw enemy states up to ~2000 ticks, oldest dropped when full - [ ] Normalization uses game constants only — no running stats, no allocations per tick - [ ] Self-check: `when isMainModule` with `doAssert` validates normalization ranges and buffer behavior - [ ] Integrated into bot skeleton's tick loop — data collection runs every tick ## Blocked by - https://git.fossellini.top/SirStone/SirRoboGarage/issues/81

SirStone added the ready-for-agent label 2026-08-25 19:11:37 +02:00

SirStone referenced this issue

2026-08-25 19:11:53 +02:00

TOPO_Gun with GF targeting #84

SirStone closed this issue

2026-08-25 19:35:07 +02:00

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: SirStone/SirRoboGarage#83