Dialogue-based Generation Of Self-driving Simulation Scenarios Using Large Language Models

Miceli-barone Antonio Valerio, Lascarides Alex, Innes Craig. Arxiv 2023

[Paper]    
Multimodal Models Reinforcement Learning Tools

Simulation is an invaluable tool for developing and evaluating controllers for self-driving cars. Current simulation frameworks are driven by highly-specialist domain specific languages, and so a natural language interface would greatly enhance usability. But there is often a gap, consisting of tacit assumptions the user is making, between a concise English utterance and the executable code that captures the user’s intent. In this paper we describe a system that addresses this issue by supporting an extended multimodal interaction: the user can follow up prior instructions with refinements or revisions, in reaction to the simulations that have been generated from their utterances so far. We use Large Language Models (LLMs) to map the user’s English utterances in this interaction into domain-specific code, and so we explore the extent to which LLMs capture the context sensitivity that’s necessary for computing the speaker’s intended message in discourse.

Similar Work