Mobile Game AI Development: Smart Opponents with BT & DDA

Your mobile shooter: the enemy sees players through walls, reacts in 0.02 seconds — players leave after a day. Sound familiar? This is a typical mistake: lack of a perception system and a naive finite state machine. Our team with 5+ years of mobile game development experience solves this in 2–4 weeks. We combine Behaviour Trees, physically honest sensors, and Dynamic Difficulty Adjustment (DDA). Developing AI opponents is one of our core competencies. The result: an enemy that feels alive but does not cheat. In 5 years on the market, we have implemented AI for 15+ mobile games, reducing player churn by 20% on average.

Why Behaviour Trees for Complex Combat?

Behaviour Trees are a hierarchical structure of Selector, Sequence, Condition, and Action nodes. They are 3–5 times easier to scale than FSM: instead of a tangled state graph, a clear tree. With 10+ behaviors, BT speeds up debugging by 2–3 times. FSM works for 3–4 states (Patrol→Chase→Attack), but the graph explodes as complexity grows. BT is the choice for complex combat.

// Unity: simplified Sequence node
public class SequenceNode : BehaviourNode {
    private List<BehaviourNode> children;
    private int currentIndex = 0;

    public override NodeStatus Tick(AIContext context) {
        while (currentIndex < children.Count) {
            var status = children[currentIndex].Tick(context);
            if (status == NodeStatus.Running) return NodeStatus.Running;
            if (status == NodeStatus.Failure) {
                currentIndex = 0;
                return NodeStatus.Failure;
            }
            currentIndex++;
        }
        currentIndex = 0;
        return NodeStatus.Success;
    }
}

We use parallel tasks for simultaneous actions (shooting + movement) and priority selectors for urgent situations (dodging a grenade). Each tick updates the Blackboard — a shared data store between nodes. This allows flexible state management without global variables.

How AI Sees the Player: Perception System

The key question is not "what does the enemy do," but "what does it know about the world." Without a proper perception system, a BT tree works with telepathy: the enemy knows the player's position through walls and reacts instantly.

Field of View — a vision cone with angle and distance:

bool CanSeePlayer(Transform enemy, Transform player, float viewAngle, float viewDistance) {
    Vector3 dirToPlayer = (player.position - enemy.position).normalized;

    if (Vector3.Angle(enemy.forward, dirToPlayer) > viewAngle / 2f)
        return false;

    float dist = Vector3.Distance(enemy.position, player.position);
    if (dist > viewDistance) return false;

    return !Physics.Raycast(enemy.position, dirToPlayer, dist,
                             LayerMask.GetMask("Obstacles"));
}

Hearing — sound events through a queue: footsteps, gunshots, falling objects. Each event has a radius and attenuation. Memory — we remember the last known position and patrol it if we lose the target.

What is DDA and How Does It Adjust Difficulty?

Fixed difficulty is bad. Dynamic Difficulty Adjustment adapts behavior based on player performance. Parameters are loaded from a config — designers can change them without recompilation.

DDA automatically shifts parameters based on the player's win rate over the last N sessions. This saves up to 30% of manual balancing time and keeps the player in the flow. Wikipedia: Dynamic game difficulty balancing

How to Ensure Performance on Low-End Devices?

AI ticks should not happen every frame. For 20 enemies on the scene:

Unity NavMesh works well on mobile, but recalculating paths for all agents in one frame causes a spike. We distribute recalculations via CoroutineManager or Job System. We guarantee stable 60 FPS on 5-year-old devices.

AI Opponent Development Process

1. Analysis — study game mechanics, design document, behavior scenarios.
2. Design — choose architecture (FSM/BT), design perception system, memory.
3. Implementation — write code, integrate with navigation (NavMesh), tune parameters.
4. Testing — playtesting: measure death rate, time-to-kill, win rate per level. Iterate.
5. Deployment — hand over source code with comments, BT diagrams, configs. Train the team.

Each stage is documented. On request — integration into your CI/CD pipeline.

What's Included in AI Opponent Development

Within the project, you receive:

• Source code of the AI system in Unity C# or Godot GDScript with comments
• Documentation on the BT architecture and perception system (diagrams, configs)
• Integration with your navigation (NavMesh) and animations
• DDA parameter tuning for your game design
• Loadable configs (JSON) for quick designer iteration
• Team training (1-2 hours) on using and extending the AI
• Support during testing and post-release

Timeline Estimates

• Basic FSM opponent with patrol and pursuit — 3–5 days.
• Full system with BT, perception, DDA, and multiple enemy types — 3–6 weeks.

Exact cost is calculated individually after analyzing your game.

Contact us to get a consultation on AI architecture for your project. We will help you choose the optimal stack (Unity, Godot, Unreal) and avoid common pitfalls — AI cheating and performance drops.

Machine Learning in Mobile Apps: CoreML, TFLite, and On-Device Models

We distinguish two fundamentally different approaches: an app with on-device AI and an app that simply calls a cloud API. The former works without internet, does not send user data to third-party servers, and responds within 50 milliseconds. The latter depends on network latency and pricing plans. Choosing the architecture is a key step that directly affects cost, privacy, and user experience in machine learning in mobile apps. Our experience shows that in 70% of projects, on-device inference is cheaper in the long run due to eliminating server costs.

How to Choose Between CoreML and TFLite for On-Device Inference?

CoreML — Apple's native framework for running ML models on device. Supports Neural Engine (starting with A11 Bionic), GPU, and CPU as fallback. Models are converted to .mlmodel format via coremltools from PyTorch, ONNX, or TensorFlow. Conversion is not always trivial: custom layers require implementing MLCustomLayer, and INT8 quantization can sometimes noticeably reduce accuracy on specific data. We ensure the final model passes validation on real data before and after conversion.

TensorFlow Lite — cross-platform alternative for Android and Flutter. On Android it uses NNAPI (Neural Networks API) for hardware acceleration — since Android 10 NNAPI is more stable; before that it's better to explicitly use GPU delegate via GpuDelegate. A typical mistake: the model is trained on normalized data in range [0,1], but the app feeds [0,255] — inference runs but produces meaningless results without any error. We include an automatic input data validation module in the SDK.

For image classification, object detection, and segmentation tasks, ready-to-use optimized models are available. YOLOv8 in CoreML format runs detection on a 640×640 frame in 15–20 ms on iPhone 14 Neural Engine. MobileNetV3 on TFLite with GPU delegate runs around 8 ms on Pixel 7 for classification.

What If You Need Text Generation On-Device?

Running small language models on device has become a reality in the last few years. Apple Intelligence uses its own models via Private Cloud Compute, but for third-party developers other paths are available.

llama.cpp with Metal backend on iOS is a working approach for phi-3-mini (3.8B parameters, 4-bit quantization, ~2.3 GB). Inference: 15–25 tokens/second on iPhone 15 Pro. For integration in Swift, use the Swift Package llama.swift or a wrapper via C interface llama.h. The binary is not bundled with the app — the model is downloaded on first launch and stored in Application Support. Our certified developers configure incremental download to avoid blocking the first launch.

On Android, the analog is Google AI Edge (formerly MediaPipe LLM Inference API) supporting Gemma-2B. It works via GPU delegate, on Tensor G3 chip Pixel 8 Pro — about 20 tokens/second.

Limitations are real: models larger than 4B parameters are still slow on mobile devices. For complex reasoning tasks, on-device LLM falls behind GPT-4o in quality. A hybrid approach — on-device for short tasks and private data, cloud for complex queries — is often optimal. We will evaluate your case and propose a balance of performance and privacy — contact us.

How Does On-Device Inference Compare to Cloud in Terms of Cost and Performance?

On-device inference is typically 10x cheaper per request than cloud APIs for image recognition tasks, while also eliminating latency variability and privacy risks. The table below summarizes the trade-offs.

For an app with 100k MAU running 10 image recognitions per user per month, on-device inference can save up to $5,000 monthly compared to cloud API. Get a free consultation on your ML architecture today.

Integrating OpenAI API and Other Cloud Models

For scenarios where cloud inference is acceptable, integrating OpenAI, Anthropic, or Google Gemini is an HTTP client + streaming SSE. In Swift, AsyncThrowingStream is convenient for streaming responses. In Kotlin, use Flow.

Critically: API keys must never be stored in the app bundle. Even an obfuscated key can be extracted from the IPA in 10 minutes using strings or frida. Correct architecture: mobile app → your own backend → OpenAI API. The backend controls rate limiting, logs requests, and protects the key.

What Is Included in the Work (Deliverables)

• Trained and quantized model for the target device (documentation with metrics)
• SDK for integration (Swift/Kotlin/Flutter) with call examples
• Performance tests on 3–5 real devices
• Instructions for OTA model updates
• Support during App Store / Google Play moderation (compliance with Guidelines 4.2, 5.1)
• 2 weeks of technical support after release

Typical Project Pipeline

1. Task analysis — measure latency, privacy, size, supported devices.
2. Model prototyping — in Python, evaluate accuracy on target data.
3. Conversion and quantization — for CoreML/TFLite with validation.
4. Integration into the app — model wrapped in a service layer (easy to swap CoreML ↔ TFLite ↔ cloud).
5. Testing — on real devices, measure FPS, RAM, battery.
6. Deployment — via TestFlight / Firebase App Distribution, monitor metrics.

Timelines: integration of a ready CoreML/TFLite model — 1–2 weeks, development of a custom model with mobile optimization — from 6 weeks, on-device LLM chat with personalization — 4–8 weeks.

Why We Take on Complex Cases?

10+ years of experience in mobile development, 50+ implemented AI/ML solutions, guarantee of compatibility with current iOS and Android versions. All projects undergo code review and load testing. The cost includes preparation of moderation documentation and training of your team.

Contact us — we will help you choose the architecture and implement ML in your app turnkey. Order an audit of your existing solution — we will assess the potential for server cost savings free of charge. In some projects, savings can reach significant amounts per month.