Robots

The Skyfall concept involves deploying six scout helicopters on Mars to explore the planet. These helicopters are designed to operate autonomously and can be used for tasks such as identifying areas with water or other resources and selecting sites for future astronaut missions.

The Skyfall concept builds upon the success of NASA's Ingenuity Mars helicopter, which was co-developed by AeroVironment and NASA's Jet Propulsion Laboratory. Ingenuity was the first aircraft to fly under its own power on another planet, demonstrating the feasibility of rotorcraft on Mars.

Cyborg beetles are equipped with removable microchip backpacks that use electrodes to control their antennae and forewings, allowing precise directional guidance via remote control, such as a video game controller. This enables operators to guide the beetles through complex environments like rubble to locate survivors.

Cyborg beetles naturally possess sophisticated sensing capabilities, soft environmental interactions, and the ability to climb vertical surfaces and maneuver through small, complex spaces like dense rubble. These traits make them more effective than similarly sized robots, which struggle with locomotion on vertical surfaces and navigating chaotic disaster environments.

The jet-powered humanoid robot, known as iRonCub3, achieves vertical takeoff by integrating jet propulsion with an advanced whole-body control architecture. It uses AI-based control systems and aerodynamic modeling to maintain dynamic stability during flight. The control pipeline includes a Model Predictive Controller (MPC) that manages thrust and balance, allowing the robot to lift approximately 50 cm off the ground while maintaining stability. This system has been validated through both simulation and real-world experiments, demonstrating accurate tracking and control during vertical takeoff.

Developing and operating a jet-powered humanoid robot involves several challenges, including managing the extreme conditions of jet propulsion such as high air temperatures (up to 700 degrees Celsius) and supersonic air speeds (around 1800 km/h). Safety protocols and experimental procedures are critical to handle these conditions. Additionally, accurately estimating thrust forces is pivotal for stable flight, as uncertainties in the robot’s base pose and thrust can affect control. The system must also handle complex dynamics, estimation errors, and external disturbances, requiring robust control algorithms and continuous improvements in modeling and sensor accuracy.

The Realbotix Robotic AI Vision System enables humanoid robots to detect human presence and dynamically adjust their facial expressions, creating emotionally engaging and natural responses that reduce the 'uncanny valley' effect. It also supports facial recognition for personalized interactions, real-time object identification, scene awareness, and integrates multimodal AI and conversational AI to provide contextually nuanced conversations.

Realbotix humanoid robots can integrate with major AI platforms such as OpenAI's ChatGPT, Meta's Llama, Google's Gemini, and DeepSeek R1. This integration allows for enhanced conversational abilities in multiple languages, including Spanish, Cantonese, Mandarin, French, and English, and supports both local and cloud-based AI applications. Additionally, Realbotix's proprietary lip sync technology ensures precise mouth movements, improving the realism and accuracy of robotic speech synchronization.

Projects like these often use servo motors, Raspberry Pi boards, and sometimes additional components like cameras or ultrasonic sensors to control and animate the robotic eyes. For example, a project might use a Raspberry Pi with a servo driver to control eye movements, and a camera with software like Google’s MediaPipe Face Mesh for eye tracking[2][5].

These projects typically involve using sensors (like ultrasonic sensors or cameras) to detect movement or track objects. The data from these sensors is then processed by a microcontroller, such as a Raspberry Pi or Raspberry Pi Pico, which controls servo motors to move the eyes in response to the detected movement[2][5].

The Redwood AI model is an advanced artificial intelligence system developed by 1X Technologies specifically for its NEO humanoid robot. Redwood enables NEO to autonomously perform household tasks such as laundry, answering doors, and navigating familiar spaces by integrating perception, movement, and interaction capabilities. The model is trained on real-world data from both EVE and NEO robots, allowing it to generalize across tasks, handle unfamiliar objects, and exhibit learned behaviors like hand selection and retrying failed grasps. Redwood also supports whole-body and multi-contact manipulation, enabling NEO to coordinate locomotion and manipulation for complex actions such as bracing or leaning during tasks. The system operates efficiently on NEO’s onboard embedded GPU and integrates with an off-board language model for real-time voice control.

Unlike traditional robotic control systems that separate locomotion and manipulation into distinct modules, Redwood AI fuses these capabilities into a unified system. This allows the NEO robot to perform whole-body and multi-contact manipulation, such as bracing against a wall while opening a heavy door or kneeling to pick up objects, which is difficult for most robots. Redwood’s embodied learning approach enables NEO to learn from real-world interactions and adapt to new environments, improving its autonomy over time. The system also supports mobile bi-manual manipulation, allowing NEO to move and manipulate objects simultaneously, and integrates with a language model for voice-based user intent prediction.