Machine Learning for Drones & Unmanned Systems

Machine Learning Technology

Machine learning is a subset of AI (artificial intelligence) technologies that aim to allow computers to perform tasks and make decisions without the need to be explicitly programmed with exact instructions. This is achieved via the use of training or sample data, and allows machines to imitate the way that human brains act, with increasing accuracy achieved over time.

Deep Learning Tools

Rajant’s Wireless Mesh Network Node for industrial machine learning applications

Deep learning is a particular kind of machine learning that requires less classification of the data by humans beforehand. Deep learning uses algorithms such as neural networks to automate feature extraction and categorisation, and can work with much more unstructured data.

Deep learning requires higher complexity and greater setup time than non-deep machine learning, but needs less supervision once set up, provides results faster, and can be applied to more complicated problems.

Unmanned Applications

Machine learning and deep learning tools can be applied to a wide variety of drone and robotics applications. A common usage is to provide UAVs (unmanned aerial vehicles) with smoother flight control, allowing the aircraft to choose movements that result in the desired outcomes and thus cutting down on operation times. Enhanced collision, obstacle and hazard avoidance, as well as intelligent drone swarm operations, can also be achieved via deep learning.

ViDAR Maritime – real time object detection by Sentient Vision

Deep learning visualization can also be applied to images and video captured by drone and unmanned systems camera payloads. The output can be analyzed to identify and track targets or potential threats for enhanced ISR (intelligence, surveillance and reconnaissance) and situational awareness, or to spot faults in wind turbines, power lines and infrastructure.

Embedded Computing Hardware

EIZO’s Condor XR1S 6U VPX card offers superior graphics and GPGPU compute capabilities for AI and deep learning

Drones and robotics may use embedded processing hardware to run machine learning and deep learning software. These embedded solutions may involve single-board computers (SBCs) that integrate general-purpose graphics processing units (GPGPUs) and central processing units (CPUs). Drones and unmanned vehicles with a low SWaP (size, weight and power) budget may also be able to offload machine and deep learning processing to remote cloud servers.