Official implementation of: A Comparison of Prompt Engineering Techniques for Task Planning and Execution in Service Robotics by Jonas Bode, Bastian Pätzold, Raphael Memmesheimer and Sven Behnke, published at Humanoids 2024.
This simulation works using Python. Needed packages are NumPy as well as the API packages for OpenAI and Mistral. These can also be installed using pip.
pip install -r .\requirements.txt
How to configure and run an experiment is showcased in example.py. During execution transcripts of the conversation with the LLM will be saved. Each trancript is named <seed number>.txt. In addition a summarizing file called meta.txt will be created in the directory which contains metrics of all samples generated in this experiment.
If you wish to get access to the transcripts generated for our paper please contact Jonas Bode per email s6jobode@uni-bonn.de.
This repository is maintained by Jonas Bode.
Please consider citing our paper if you find our work or our repository helpful. Bibtex will follow here upon publication.
