Mirokaï

Mirokaï, developed by French robotics company Enchanted Tools founded in 2021 by Jérôme Monceaux (creator of Nao and Pepper at Aldebaran Robotics) and Samuel Benveniste, represents a paradigm shift in social humanoid robotics. With over €17 million in seed funding—the largest for French robotics—and a team exceeding 100 engineers in Paris, the company aims to deploy 100,000 units globally within a decade, targeting healthcare, hospitality, events, and education. Mirokaï's architecture uniquely combines a non-legged ball-bot base with a humanoid torso, enabling omnidirectional mobility without the complexity of bipedal locomotion. The rolling globe system, actuated by omnidirectional wheels tested for 5,000 km durability, provides 3 DoF at the base, achieving speeds up to 3.2 km/h while maintaining stability in dynamic human environments via VSLAM (Visual Simultaneous Localization and Mapping) with social awareness protocols. The upper body features 26 degrees of freedom overall: 3 DoF neck, 4 DoF per arm (shoulder, elbow), 3 DoF per wrist, 2 DoF per hand (with 8 fingers total), and 1 DoF per ear for expressive gestures. Compliant, torque-controlled arms use high-speed brushless motors (maxon drives) with high torque density (2x faster in ES model), enabling precise manipulation up to 1.5 kg payload. Hands incorporate Hall effect and contact sensors for AI-driven grasping, recently upgraded for reliability. Perception is robust: 2 RGBD cameras (1920x1080), 2 infrared, 9 time-of-flight (ToF), 6 ultrasonic sensors, 4 microphones, 8 torque sensors, 3 IMUs, and Bluetooth receiver, ensuring 360° coverage and GDPR-compliant face tracking. Compute relies on 2x NVIDIA Jetson NX modules (upgraded to state-of-the-art NVIDIA GPU in Explorer Suit), running a custom Linux OS for low-latency processing. AI architecture employs a multi-AI stack: agnostic multi-LLM integration for fluid, multilingual conversations with emotional prosody; advanced vision-language models (VLM) for context-aware semantic understanding; real-time speech recognition/synthesis; and agentic behaviors for autonomous task execution like logistics (tray carrying, mail delivery) without user training. The face, powered by Unity 3D animations, features a fully modeled 3D head with teeth, tongue, and ultra-smooth transitions (laugh, wink, grumpy), syncing emotions to arms for 'emotivation by design.' A developer API and simulator facilitate customization. Deployments highlight real-world viability: First commercial unit to ISIR/Sorbonne (July 2024) for HRI research; AP-HP Broca Hospital (France) for patient assistance; Institut du Cancer de Montpellier for pediatric oncology support; Oulu University (Finland) for care/rehab/education; 11 units in APREH assisted living; CES/Vivatech demos; Kanematsu Japan partnership (2025). Safety features include RING system, compliant actuators, and human-safe design (IP unspecified). Battery: Saft Li-ion packs for 4-8 hour shifts, 1-hour recharge. Priced ~$30k upfront + maintenance, Mirokaï transitions from prototype to B2B scale by 2026, emphasizing 'enchantment' via storytelling and emotional design to foster trust and adoption.