ai

AI taking part in a game of Battle of the Centerländ.

Structure

This project contains two programs: ai and ai_train. ai is the actual program implementing the AI. ai_train is a wrapper used to train the model for the AI.

Usage

AI

The AI takes the following arguments:

Short	Long	Environment	Description	Default
-t	--host	HOST	Host to connect to.
-p	--port	PORT	Port of host to connect to.	3018
-n	--name	NAME	Name to use in game.
-d	--delay	DELAY	Delay in ms to wait before declaring as ready to the server.
-m	--mode	MODE	AI Mode for card choice, on of trained or random.	trained
-o	--model-path	MODEL_PATH	Model file to use in the trained mode. Ignored in other modes.	/var/lib/app/model/forest.model
-e	--exploration-rate	EXPLORATION_RATE	Exploration rate to use in the trained mode. Ignored in other modes.
-c	--compatible	COMPATIBLE	Enable compatibility mode, where the client will wait before sending messages.	false
-r	--reconnect-tries	RECONNECT_TRIES	Maximum times to try a reconnection.	10
-l	--log-level	LOG_LEVEL	Logging level, one of trace, debug, info, warn or error.	info

Run ai --help to see the full help message.

Training Wrapper

The training wrapper takes the following arguments:

Short	Long	Environment	Description	Default
-o	--model-path	MODEL_PATH	File to save the model to.	forest.model
-i	--iteration	ITERATIONS	Exploration rate to use in the trained mode. Ignored in other modes.	10
-g	--games	GAMES	Enable compatibility mode, where the client will wait before sending messages.	100
-s	--server-path	SERVER_PATH	Path to server binary.	/usr/bin/server
-a	--ai-path	AI_PATH	Path to ai binary.	/usr/bin/ai
-l	--log-level	LOG_LEVEL	Logging level, one of trace, debug, info, warn or error.	info

Run ai_train --help to see the full help message.

Podman/Docker

To run the AI via Podman or Docker, use

podman run \
    --name AI1
    --delay 1000
    --host <host>
    gitlab.uni-ulm.de:5050/softwaregrundprojekt/2022-2023/gruppenprojekt/group-03/ai-03

This will:

• connect to <host>:3018
• wait for 1 second
• log in as AI1

Note: You might need to log in first:

podman login -u <username> -p <access token> gitlab.uni-ulm.de:5050

Build

Prerequisites

• Rust
  • Cargo
  • clippy (for code analysis)
  • rustfmt (for code formatting)

Everything can be installed with rustup:

# Update rustup's information.
rustup update

# Install the stable Rust toolchain (if not already present).
rustup toolchain add stable

# Add rustfmt.
rustup component add rustfmt

# Add clippy.
rustup component add clippy

Building

To run the binary:

cargo run

To build both binaries:

cargo build

The resulting binary will be under target/debug/ai.

To build the binary in release mode:

cargo build --release

The resulting binary will be under target/release/ai.

Testing

Code Formatting

This should not output anything when the code is formatted corretcly:

cargo fmt --check

Run the following to automatically format the Rust code:

cargo fmt

Code Analysis

Run the following to check the code for common mistakes:

cargo clippy

Tests

Run tests with the following:

cargo test