Abstract

Robotics has made remarkable hardware strides-from DARPA's Urban and RoboticsChallenges to the first humanoid-robot kickboxing tournament-yet commercialautonomy still lags behind progress in machine learning. A major bottleneck issoftware: current robot stacks demand steep learning curves, low-level C/C++expertise, fragmented tooling, and intricate hardware integration, in starkcontrast to the Python-centric, well-documented ecosystems that propelledmodern AI. We introduce ARK, an open-source, Python-first robotics frameworkdesigned to close that gap. ARK presents a Gym-style environment interface thatallows users to collect data, preprocess it, and train policies usingstate-of-the-art imitation-learning algorithms (e.g., ACT, Diffusion Policy)while seamlessly toggling between high-fidelity simulation and physical robots.A lightweight client-server architecture provides networkedpublisher-subscriber communication, and optional C/C++ bindings ensurereal-time performance when needed. ARK ships with reusable modules for control,SLAM, motion planning, system identification, and visualization, along withnative ROS interoperability. Comprehensive documentation and case studies-frommanipulation to mobile navigation-demonstrate rapid prototyping, effortlesshardware swapping, and end-to-end pipelines that rival the convenience ofmainstream machine-learning workflows. By unifying robotics and AI practicesunder a common Python umbrella, ARK lowers entry barriers and acceleratesresearch and commercial deployment of autonomous robots.