The SharpaWave is a Component humanoid robot developed by Sharpa (Sharpa Robotics). It features 22 degrees of freedom.
Primary use case: Autonomous food-service manipulation, including preparing and serving ice cream
DOF
22
Degrees of Freedom
22
Locomotion
stationary
Hand Type
Tactile-sensitive robotic hand
Max Lift/Arm
6 kg
AI Capabilities
Next-physical-state prediction, Vision-language-action trajectory control, Real-time tactile-based grasp control, Continuous force control, Embodied short-term memory, Adaptation to real-world variability, Autonomous contact-rich manipulation
Primary Use Case
Autonomous food-service manipulation, including preparing and serving ice cream
Rhoda AI demonstrated a robot searching for and retrieving textured balls with its head and wrist cameras blocked, relying entirely on touch to guide its movements. The system combines a new tactile sensor with Rhoda's Direct Video-Action model, which predicts future tactile signals that an inverse dynamics model then converts into robot actions. The article notes the demonstration is illustrative but lacks repeated-trial results or a reported success rate, leaving reliability unquantified.
Markov Robotics uses LTX-2.5 as a video-based manipulation policy, generating video of task completion from head-cam footage and executing what the video implies. The company selected LTX for its token efficiency, enabling closed-loop inference on their own hardware, with fine-tuning reportedly taking under an hour to predict robot action tokens alongside video tokens.
Google DeepMind released Gemini Robotics 2, the first AI model suite providing whole-body control for humanoid robots by splitting vision-language-action, agentic reasoning, and on-device execution into three distinct models. The release coincided with the FCC's ban on Chinese humanoid robot imports, positioning Google's software as the standard for domestic humanoid autonomy while creating a competitive landscape where robot intelligence outweighs hardware chassis value. Partners including Apptronik, Boston Dynamics, and Agile Robots are already integrating these capabilities into high-stakes environments.