Kepler

The Kepler Forerunner series, developed by Shanghai Kepler Exploration Robot Co., Ltd. (founded in 2023), represents a rapid advancement in Chinese humanoid robotics, positioning itself as a competitor to Tesla's Optimus. The series began with the Forerunner K1 prototype unveiled in late 2023, debuting at CES 2024, featuring a highly bionic structure with 40 degrees of freedom (DOF), 178 cm height, 85 kg weight, and 100 TOPS computing power via a self-developed Nebula system. This system integrates visual SLAM for 3D modeling, multimodal interaction (language, vision), hand-eye coordination, and dynamic sensing for environmental adaptation. Propulsion relies on proprietary KEPLERGEAR actuators: planetary roller screw actuators in limbs delivering up to 8,000 N thrust for high precision and power, paired with rotary actuators (peak 200 N.m torque, 0.01° accuracy) for shoulders and waist. Hands mimic human dexterity with 12 DOF, force sensing, and millimeter-level control. Evolving to Gen 5.0, the Forerunner K2 (2024) boosts to 52 DOF (including 2 head DOF for natural interaction, 11 DOF per hand), with rope-driven tactile manipulators (15 kg single-hand payload, 25 force touchpoints per finger, 96 contact points total per hand in some configs). It features a 2.33 kWh high-capacity battery enabling 8-hour continuous operation, validated in an industry-first 8-hour livestream at WAIC 2025 as 'Bumblebee' variant (175 cm, 75 kg, 30 kg payload). Architecture employs series-parallel actuators for energy efficiency (near-zero idle consumption), star-shaped wiring for maintenance ease, and tri-component motion system: embodied perception (multi-camera, sensor arrays), decision-making (cloud-based cognitive models), and execution. AI leverages imitation/reinforcement learning, micro-models distilled from large foundation models (partnered with Chinese LLMs) for real-time perception (auditory, tactile, force, positional), gait planning (improved stability/speed), task planning, and full-body coordination. Nebula evolves with cerebellar balance control mimicking human reflexes. Real-world deployments focus on blue-collar tasks: intelligent manufacturing (assembly, inspection), warehousing/logistics (sorting, handling deformable objects), high-risk operations, and education/research. K2 underwent customer trials for material handling, stamping, patrol, and quality checks; WAIC 2025 demos included logistics sorting, automotive manufacturing, interactive games. Commercialization progresses in phases: initial pilots, vertical generalization, universal apps, with small-batch production and pre-sales (starting ¥248,000 for developer configs). Emphasizing mass production over stunts, Kepler collaborates with 50+ customers, targeting high reliability, safety, and cost-effectiveness (~$30,000 est. for earlier models). Sensors include wide-angle binocular cameras, 4-mic arrays, accelerometers, AHRS, 6-axis force wrists, dynamic force/tactile arrays. No specific CPU models disclosed, but 100+ TOPS edge/cloud hybrid enables low-latency embodied intelligence. Challenges addressed: endurance, rigidity, interactivity via optimized algorithms. As of late 2025, Kepler accelerates toward industrial scalability, with Bumblebee proving shift-long viability.