Improving Translation of English Instructions With Sequential Constraints to Linear Temporal Logic in Navigation Tasks via Prompt Engineering

Abstract

Autonomous agents like self-driving cars, robots, and drones are becoming more relevant and prevalent in society. It is an immediate advancement of technology with many benefits to society because of how applicable it is. My goal is to enable autonomous agents to be more accessible and widely deployed across cities, focusing on robots executing navigation tasks in cities. I do this by researching methods to improve translation accuracy from English instructions with sequential constraints to Linear Temporal Logic (LTL) and improve generalizability in this translation domain. I take an LSTM-based approach and a large language model approach, focusing on prompt engineering in the latter.