Test-Driven Agentic Framework for Reliable Robot Controller

TL;DR

This paper introduces a test-driven, agentic framework for synthesizing reliable robot controllers that iteratively improve through diagnostic feedback, enhancing robustness in navigation tasks within 2D and 3D environments.

Contribution

The work presents a novel dual-tier repair strategy for controller synthesis that combines prompt-level refinement with direct code editing, improving reliability over traditional methods.

Findings

Test-driven synthesis significantly improves controller robustness.

Framework performs well in both 2D and 3D navigation tasks.

Iterative refinement outperforms one-shot controller generation.

Abstract

In this work, we present a test-driven, agentic framework for synthesizing a deployable low-level robot controller for navigation tasks. Given a 2D map with an image of an ultrasonic sensor-based robot, or a 3D robotic simulation environment, our framework iteratively refines the generated controller code using diagnostic feedback from structured test suites to achieve task success. We propose a dual-tier repair strategy to refine the generated code that alternates between prompt-level refinement and direct code editing. We evaluate the approach across 2D navigation tasks and 3D navigation in the Webots simulator. Experimental results show that test-driven synthesis substantially improves controller reliability and robustness over one-shot controller generation, especially when the initial prompt is underspecified. The source code and demonstration videos are available at: https://shivanshutripath.github.io/robotic_controller.github.io.