Drones Are Able to Navigate in Unknown Environments Using Fluid Neural Networks
Researchers at the Massachusetts Institute of Technology (MIT) believe that thanks to liquid neural networks, they have found a more efficient way to help drones fly in uncharted space.
Fluid neural networks are able to continuously adapt to input data and make the right decisions about choosing a path. This will help you navigate unknown and challenging terrains such as forests, cities, and noisy environments. These more advanced drones will be better able to do search and rescue, rescuing people, delivering things, and observing nature.
A new class of machine learning algorithms extracts a causal structure from multidimensional unstructured data received from drone cameras. These algorithms can understand the task assigned to them and discard everything unnecessary. Such developments will allow the drone to work in different conditions without the need to train them for each environment.
Experiments have shown that it is possible to teach a drone to find an object in a summer forest, and then use it in winter in a completely different environment. Such adaptability is based on a causal relationship.