FEATURED ROBOTICS

AI Learns by Simulating the World

From arXiv • Latest Research

Photo by Tim Cooper on Unsplash • Unsplash

We've created an artificial intelligence system that can perform many tasks at once by imagining and simulating different scenarios,.

Read Full Paper →

Robots Learn to Adapt Tasks with AI-Powered Generalization

Scientists taught robots to perform tasks by moving them around with their hands, which simplified the process of teaching robots new skills. This method allows robots to adapt and learn to perform tasks in a variety of situations, such as cleaning different surfaces or dressing people in various arm postures, by learning from these initial demonstrations.

Read Paper →

AI Learns from Data to Perfect Nuclear Plant Operations

Scientists created a detailed, simulated model of a nuclear power plant that can be used to test and improve how robots work in these environments. This digital twin has the potential to greatly benefit the nuclear industry by allowing for more efficient monitoring, maintenance, and operation of power plants, which could lead to improved safety and reduced costs.

Read Paper →

Photo by Ed 259 on Unsplash • Unsplash

AI-Powered Underwater Robots Get Smarter with Every Dive

Scientists developed a new system called UROSA that helps underwater robots make decisions quickly and accurately in unpredictable environments. This innovation has the potential to greatly improve the way underwater vehicles navigate and complete tasks, making them safer and more effective for a variety of real-world applications.

Read Paper →

AI Learns to Pick Objects Safely and Efficiently

We created a large database of images with labels showing what objects can be grasped by robots, along with instructions for how to grasp them in different situations. This new tool could help robots learn to pick up objects more efficiently and safely in various environments, which is important for making robotics technology more useful and reliable.

Read Paper →

AI Learns to Feel Its Way Through Complex Spaces Like a Cockroach

Scientists created a small, lightweight, and low-power sensor that mimics the way cockroaches use their antennae to feel their surroundings. This new sensor could potentially enable robots as small as insects to navigate complex spaces more effectively by sensing obstacles and textures, which would be important for tasks like search and rescue operations in confined areas.

Read Paper →

AI Learns to Fix Corrupted Images for Smarter Self-Driving Cars

Scientists created a system to fix corrupted images from cameras before they affect how a car's control systems work. This new method can improve a self-driving car's ability to navigate in different weather conditions and lighting, making it more reliable and reducing errors caused by poor image quality.

Read Paper →

Robots Get a Grip with AI-Powered Object Recognition

Scientists created a new system for robots to pick up objects, called KGN-Pro, which can use 3D information from cameras to make more accurate guesses about what the robot should grab. This innovation has the potential to improve robotic grasping abilities and could be useful in various industries where robots interact with objects.

Read Paper →