Key Technologies of the 3 Types of Robots

Nikola

• Reproduction of human facial muscle movements: Actuators (pneumatically driven "motors") are designed based on knowledge of psychology and anatomy to replicate the movements of human facial muscles. A world-leading 35 of these actuators are implemented in a life-sized human head.
• Emotionally expressive capabilities recognizable by humans: Demonstrated, Groundbreaking research in human-recognizable robot emotions. In 2022, we initially announced the "six basic emotions" (simple emotions such as anger and surprise). (News release on February 10, 2022: "Development of an android capable of facial expressions like a human"). In 2025, 13 "complex emotions" were also announced (January 24, 2025: "Paper by former visiting researcher Diel et al. accepted by Scientific Reports").
• Eye movements: Achieves natural, human-like eye movements by modeling gaze behavior dependent on interaction situations and models based on individual human personalities.
• Head movements: Implemented a system that generates robot head nodding movements in real-time, synchronized with the robot's speech.
• Mouth movements: Implemented a system that generates robot lip movements in real-time, synchronized with the robot's speech.

Aetro

• Autonomous assistance: Estimates the wearer's motion intent using myoelectricity and camera information, and the robot generates support movements according to human motion.
• Motion assistance by pneumatic artificial muscles: Achieves weight reduction while ensuring operational force by using pneumatic artificial muscles instead of hydraulic or electric motors. Furthermore, when uncontrolled, the joints can be easily moved by external force (in the case of motor drive, it is usually difficult to easily move them by external force when uncontrolled).
• Lightweight assistive device: Uses a lightweight frame for the motion assistance device and achieves lighter weight in the wearable parts through drive by pneumatic artificial muscles.
• Assistance for agile movements: Instantly generates motion assistance by appropriately combining pre-derived control laws according to the intent estimation results.

Indy

• Integration of elemental technologies aiming for domestic tasks: To be suitable for movement in human living environments and to realize natural communication with humans, the following functions are implemented in a child-sized body.
• Speech recognition corresponding to multiple sound sources: By combining voice data from a microphone array (distributed compound microphone) and image data from LiDAR (laser radar) or an external camera, they accurately recognize separated voices even when multiple speakers (multiple sound sources) speak simultaneously. We tune the language model for speech recognition system according to the application.
• Image recognition including object relationships: Using a visible light+distance measuring camera, they instantly analyze the relationships between detected objects (e.g., positional relationships, semantic relationships) and understand the situation (scene). For example, they recognize human states such as posture and gestures and estimate objects that should be focused on.
• Memory: They systematically store and organize their memories as narratives (stories). They use the narratives to select actions tailored to individual people and situations.
• Context-aware response: Based on the open and Japanese-proficient language model "llm-jp-3-13b," optimization (fine-tuning) for inference and response tailored to the situation is performed. Achieves highly accurate and real-time context-aware responses. (September 25, 2024: "Release of LLM-jp-3 1.8B, 3.7B, 13B" - National Institute of Informatics, Research and Development Center for Large Language Models).
• Speeding up and miniaturization: An externally placed system with memory controls their actions without depending on an external network and communicates wirelessly with the robots. This miniaturizes the computer inside the robot body and speeds up responses.
• Implementation of gestures: We experimentally implement gestures that are indispensable for communication in human living environments.
• Implementation of facial expressions: They have facial expressions indispensable for natural communication with humans.
• Large-capacity storage battery: It realizes spontaneous long-term activities in human living environments.
• Non-toppling property (Prototype version characteristic): Adopting an upper body design that utilizes gravity like a pendulum realizes a stable and energy-saving self-restoring mechanism and compliance (soft responsiveness) in the upper body. Reduces impact in physical contact (collision) with humans and enables soft physical interaction.