IR-SIM: A Lightweight Skill-Native Simulator for Navigation, Learning, and Benchmarking

A new lightweight robotics simulator called IR-SIM has been proposed to speed up automated navigation research by replacing complex coding with simple YAML configuration files, according to a paper posted to the arXiv preprint repository on June 7, 2026 [1][2]. The Intelligent Robot Simulator, or IR-SIM, is designed as a “skill-native” tool for rapid scenario construction, benchmarking, and robot learning [1][2]. Its creators argue that existing simulators often require custom code or intricate interfaces, which slows down the kind of automated algorithm development increasingly driven by large language models [2]. Large language models are neural networks trained on vast text corpora to generate, summarize, and analyze language, and they underpin modern chatbots [8]. In IR-SIM, every scenario is defined entirely through YAML configuration files that specify mobile robot kinematics, geometric collision checking, LiDAR sensing, visualization, and behavior modules [1][2]. The paper states that this approach makes robotic simulation “fully describable and reproducible,” enabling scenarios to be generated and modified from text prompts via proposed IR-SIM agent skills [2]. The resulting scenarios can be used for automated benchmarking of navigation algorithms and for automated generation of training data for learning methods [1][2]. The simulator also provides bridges to high-fidelity simulators and real-world deployment, allowing users to validate algorithms in more realistic settings after prototyping without writing additional code [1][2]. The experiments described in the paper cover multiple tasks: constructing navigation scenarios from natural language, training a collision avoidance policy, benchmarking social navigation policies, and bridging to high-fidelity simulators and real-world deployment [2]. The paper appeared on arXiv, an open-access repository of electronic preprints that is moderated but not peer-reviewed [6]. arXiv was founded in 1991 and now receives about 24,000 submissions per month [6]. The repository hosts a framework called arXivLabs, launched in 2020, which allows community collaborators to build experimental tools that appear on article record pages [5]. Current arXivLabs projects include the Bibliographic Explorer, which maps citation trees, and the CORE Recommender, which surfaces related open-access papers from a global network of repositories [4][5]. The IR-SIM paper’s abstract page displays several of these Labs integrations, including links to code-finding services and bibliographic tools [1].