cs.SENov 16, 2026

EmbedGym: Letting Coding Agents Develop and Debug Embedded Firmware Without Physical Hardware

Authors: Minghui Zhao, Yiming Li, Sophie Zhu, Maxwell Khor, Lauren Ouyang, Hai "Helen" Li, Yiran Chen, Tingjun Chen

Organizations: Department of Electrical and Computer Engineering, Duke University · Tesla STEM High School

Abstract

Modern coding agents perform well on general software benchmarks because they can run code and unit tests in a closed loop. Embedded firmware development, however, breaks this loop: a program that compiles can still fail on real hardware due to pin polarity, bus protocol, timing, or networking bugs that an agent at a text terminal cannot observe. Existing embedded coding agents either skip hardware execution, drive GPIO-only simulators with multi-second delays, or test on physical boards in a static environment. In this paper, we present EmbedGym, an interactive virtual hardware environment that lets coding agents develop and debug embedded firmware without physical hardware. EmbedGym uses instruction-level microcontroller unit (MCU) emulators to execute the same binary that would be flashed onto a physical board, connects them to 41 peripheral models, an analog circuit simulator, and virtual WiFi/Bluetooth Low Energy (BLE) endpoints, and exposes an interface for agents to change sensor inputs, press buttons, and inspect hardware states. On top of it, we build EmbedGym-Bench, which consists of 598 tasks across 6 MCU-framework pairs graded by hidden timed test vectors, using a five-stage multi-agent workflow that synthesizes peripheral models and tasks from datasheets and driver libraries. We expose EmbedGym to a coding agent at three levels and evaluate eight models over more than 14K trajectories. Even when compilation succeeds 99.7% of the time, 19.9% of compiled firmwares fail at runtime (up to 73% on individual targets). In contrast, interactive hardware testing in EmbedGym increases pass rates across models by +4.9% to +14.7% (up to +31% on individual targets). Our results also show that an agent with EmbedGym outperforms prior embedded agent frameworks, and lightweight models with EmbedGym match frontier models without it at 30-70x lower cost.

Figures & tables

Explore similar work

Oct 8, 2026

Closed-loop evaluation of LLM agents for embedded software development

Large language models (LLMs) are increasingly deployed as coding agents that edit files, run builds and tests, inspect execution results, and repair software iteratively. Embedded firmware is a demanding target because correctness depends on closed-loop behavior under sensing, timing, and safety constraints, not only on static source quality. Yet embedded-agent evaluation remains limited and often emphasizes one-shot synthesis or offline correctness. We present a benchmark for closed-loop evaluation of embedded coding agents. Each task provides a plain-text engineering description, constrained workspace, and visible build-and-runtime surface. The agent must translate requirements into implementation and self-verification steps, then iterate until the required device behavior is achieved. The suite contains five embedded-control tasks and four feedback scenarios: one-shot generation, realistic self-verification, CI-style red/green feedback, and oracle-style detailed feedback. The implementation targets simulated ESP32 firmware for reproducibility. We evaluate seven GPT-family and Qwen-family configurations across five tasks and four scenarios, with three repetitions per condition for 420 runs. gpt-5.4 has the highest pass rate among evaluated configurations but does not saturate the benchmark; qwen3.5-27B is the strongest observed local model; and smaller local models degrade sharply in pass rate and search efficiency. These results suggest that capable local embedded coding agents are emerging.
Oct 16, 2026

LoopHIL: Looping Hardware into LLM-assisted Agents for Wireless Embedded Systems

LLM agents are widely used for software development and debugging, but their adoption in embedded systems, especially for wireless tasks, remains limited. This is because embedded development requires hardware-in-the-loop testing. Agents cannot directly observe the spectrum, probe pins, or measure power, so they cannot verify wireless link functionality; currently, human engineers perform this verification. We systematically evaluate whether LLM agents can establish wireless links when provided with hardware access. We introduce LoopHIL, a framework that lets agents control a multi-device RF bench, monitor it with a spectrum analyzer, logic analyzer, and energy meter, and validate links against a verification ladder, all within code-enforced safety limits. Across four increasingly complex tasks, from generating a carrier to backscatter decoding by a commercial receiver, and using six models from two vendors, hardware-in-the-loop increases task completion rates from 54% to 79% and from 17% to 83% on the most challenging task. It also halves task time and output tokens. With debugging tools, agents decrease tag power consumption by up to 96%. However, small models often report unverified claims, so independent hardware checks remain necessary.
Mar 20, 2026

Skilled AI Agents for Embedded and IoT Systems Development

Large language models (LLMs) and agentic systems have shown promise for automated software development, but applying them to hardware-in-the-loop (HIL) embedded and Internet-of-Things (IoT) systems remains challenging due to the tight coupling between software logic and physical hardware behavior. Code that compiles successfully may still fail when deployed on real devices because of timing constraints, peripheral initialization requirements, or hardware-specific behaviors. To address this challenge, we introduce a skills-based agentic framework for HIL embedded development together with IoT-SkillsBench, a benchmark designed to systematically evaluate AI agents in real embedded programming environments. IoT-SkillsBench spans three representative embedded platforms, 23 peripherals, and 42 tasks across three difficulty levels, where each task is evaluated under three agent configurations (no-skills, LLM-generated skills, and human-expert skills) and validated through real hardware execution. Across 378 hardware validated experiments, we show that concise human-expert skills with structured expert knowledge enable near-perfect success rates across platforms.