v1.0.0 – Depthrun

Depthrun v1.0.0 – Pre-Defense Build

First feature-complete release of the diploma project. The custom Adaptive Enemy Behavior module is fully wired into the game and ready for committee demonstration.

W, A, S, D - Movement
LMB – Attack
1 – Sword
2 – Bow
E – Use Potion

F3 – Open DebugWidget
Y – God Mode
U – Super Attack

Highlights

Adaptive Behavior Module (core diploma deliverable)

• 3-layer decision pipeline in 100% C++ – no Behavior Tree, no ML
• Layer 1: ContextEvaluator – 6 normalized factors (distance, weapon, HP, allies, room density, agressive memory)
• Layer 2: ThreatCalculator – T_final = clamp(Σ wᵢ × fᵢ, 0, 1)
• Layer 3: StateTransitionResolver – FinalScore = Utility - Cost + Inertia + Pattern
• DynamicWeightManager adjusts weights via deterministic feedback loop
• PatternRecognizer – 2-gram + 3-gram window of size 15
• Evaluation on FTimerManager every 0.3 s
• bAdaptiveEnabled flag for A/B testing against vanilla FSM
• Real-time UDebugAdaptiveWidget overlay for committee demonstration

Gameplay Systems

• Procedural dungeon (multi-cell rooms: 1×1) - to do (1×1, 2×1, 1×2, 2×2, L-shape)
• Two-slot weapon system: Sword + Bow
• 7 per-run item modifiers (Ricochet, Pierce, Multishot, etc.)
• 4 persistent Hub upgrades (Damage, Range, ArrowCount, MaxHP)
• Chest loot system with diamonds + potions + 70% run-item drop
• Boss-room
• Niagara VFX for chest opens, victory and defeat screens
• Crossfading music system with combat trigger

Persistence

• SQLite-backed save subsystem (Saved/Depthrun.db)
• player_profile table for metaprogression
• run_history table written every run for diploma statistics

Tooling

• Real-time debug widget per enemy (max 4)
• Console commands for economy, items, DB inspection

Requirements

• Windows 10/11 x64
• Unreal Engine 5.7
• Visual Studio 2022 (Game development with C++)

Build

1. Generate VS project files from Depthrun.uproject
2. Build configuration: Development Editor | Win64
3. Open L_MainMenu and press Play

---

The game environment is a demonstration polygon. The engineering result is a software module of adaptive behavior implementing a 3-layer threat assessment model with dynamic weight correction, time-decay memory, N-gram pattern recognition, and utility-based state selection.