Crossout Automation: Distributed Bot System

This project is a game automation system for Crossout, consisting of a high-speed client and a dedicated pathfinding server. Explores real-time computer vision, multi-threaded architecture, and distributed systems.

A screenshot of the bot’s interface and processing capabilities during early-stage development.

Project Overview

The system is split into two main components to optimize performance:

1. client: Handles real-time screen capture, player location, enemy detection, game state awareness, and input simulation.
2. server: A dedicated backend that manages map data and performs pathfinding calculations.

Technical Stack

• System: Windows 10
• Language: Python 3.11
• Computer Vision: OpenCV, NumPy
• Backend & API: FastAPI, Uvicorn, WebSockets
• Automation: PyDirectInput, Win32API, ctypes
• Algorithms: Custom A* with path smoothing, ROI optimization

Core Features

• Real-Time Computer Vision: Uses OpenCV for fast detection of enemy markers and player localization on the minimap.
• Distributed Navigation: Offloads pathfinding to a FastAPI-based server to maintain high frame rates on the client.
• Multi-Threaded Execution: The client runs 6+ concurrent threads to handle independent tasks like camera input and target tracking.
• Intelligent Menu Navigation: Automatically recognizes and navigates game menus by analyzing UI templates.
• Custom Mapping Tools: Includes a system for recording and generating drivable masks from in-game data.

Lessons

• Initial Design: Getting the code foundation right at the start is essential. I had to redo the core structure multiple times.
• Language Selection: High-frequency image processing in native Python faces performance bottlenecks. For higher FPS, moving image processing tasks to C++ or Rust would be more effective.
• Distributed Design: Separating navigation from vision significantly improved client-side responsiveness.
• Custom Tooling: Building my own visualization system boosted my productivity and made algorithm outputs easier to understand.
• Input Latency: Simulating mouse and keyboard inputs introduced a 4.5ms delay, which hindered real-time performance. Moving to controller simulation would likely reduce latency and enable much smoother movement.

Project Status

Built for educational purposes, no longer being updated.