Decide debug overlay layout and content #151

New Issue

2026-09-12T11:37:17+02:00

SirStone commented

2026-09-12 11:37:17 +02:00

Question

Given the available drawing primitives (from the debug drawing API research), what should the debug overlay show and how should it be laid out on the battlefield?

Candidates:

Input layer: 36 population neurons, highlight which band is active
Hidden layer: 6 neurons, show membrane potential as fill level
Output: target angle vs actual gun direction vs enemy direction
Error value (numeric or bar)
Weight magnitudes (connection thickness?)
Learning progress (error over time)

Which of these are essential for the proof-of-concept, and what's the visual layout?

Parent map: #147

Blocked by: #148

## Question Given the available drawing primitives (from the debug drawing API research), what should the debug overlay show and how should it be laid out on the battlefield? Candidates: - Input layer: 36 population neurons, highlight which band is active - Hidden layer: 6 neurons, show membrane potential as fill level - Output: target angle vs actual gun direction vs enemy direction - Error value (numeric or bar) - Weight magnitudes (connection thickness?) - Learning progress (error over time) Which of these are essential for the proof-of-concept, and what's the visual layout? Parent map: #147 **Blocked by:** #148

SirStone commented

2026-09-12 12:19:59 +02:00

Resolution

Elements to display:

Gun direction line (from bot, showing actual gun heading)
SNN target angle line (from bot, showing where SNN wants to aim)
Enemy direction line (from bot, toward enemy)
Active input neuron(s) — which population band(s) are firing
Hidden neuron potentials — 6 bars showing membrane voltage
Output neuron potential — membrane voltage driving the target angle
Weight magnitudes — all 222 connections, alpha-coded (opacity = weight magnitude). Near-zero weights invisible, strong weights solid. Shows learning paths emerge over time.

Layout: Mixed.

Aim lines (gun direction, SNN target, enemy bearing) radiate from bot position — they're spatial.
Neuron states (input bands, hidden bars, output bar) and weight connections drawn in a fixed corner panel — they're abstract data.

Error logging: Per-evaluation to stdout: tick=N error=12.3° reward=0.075. One line per wait-then-evaluate cycle. No on-screen error display — stdout is sufficient.

## Resolution **Elements to display:** - Gun direction line (from bot, showing actual gun heading) - SNN target angle line (from bot, showing where SNN wants to aim) - Enemy direction line (from bot, toward enemy) - Active input neuron(s) — which population band(s) are firing - Hidden neuron potentials — 6 bars showing membrane voltage - Output neuron potential — membrane voltage driving the target angle - Weight magnitudes — all 222 connections, alpha-coded (opacity = weight magnitude). Near-zero weights invisible, strong weights solid. Shows learning paths emerge over time. **Layout:** Mixed. - Aim lines (gun direction, SNN target, enemy bearing) radiate from bot position — they're spatial. - Neuron states (input bands, hidden bars, output bar) and weight connections drawn in a fixed corner panel — they're abstract data. **Error logging:** Per-evaluation to stdout: `tick=N error=12.3° reward=0.075`. One line per wait-then-evaluate cycle. No on-screen error display — stdout is sufficient.

SirStone closed this issue

2026-09-12 12:20:02 +02:00

SirStone referenced this issue

2026-09-12 12:20:17 +02:00

SNNBot: SNN-aimed gun against stationary target #147

SirStone added the wayfinder:grilling label 2026-09-12 12:22:03 +02:00

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: SirStone/SirRoboGarage#151