airship_planner

This package implements a task planner leveraging LLMs. It supports both using OpenAI’s API to access GPT modules and utilizing the Ollam API to access open-source LLMs, such as LLAMA.

airship_planner API Overview

llm_planner_node

• Default configuration file: airship_planner/config/config.yaml
  • semantic_map_file: Specifies the file of the semantic file
  • llm_server_url: Specifies the url of the LLM service constructed by the user using Ollama.

1. Service
   • /airship_planner/planner_server: Provides a service for receiving user instructions, utilizing LLMs to process those instructions, and generating a list of tasks composed of navigation, grasping, and placing.
2. Parameters
   • semantic_map_dir: The directory containing the semantic map. Users can specify this parameter through the launch file.
   • semantic_map_file: The file name of the semantic map. Users can specify this parameter through config.yaml.
   • llm_server_url: The URL of LLM server. User can specify this parameter through config.yaml.

To-Do List

• Add more task primitives to the task planner and implement these new tasks in Airship.
• Use VLM models to enhance the capbility of task planning and scheduling.
• Develop a foundation world model for general embodied AI applications.