We are strong proponents of the payload autonomy paradigm, in which high-level deliberative autonomy is implemented independently of low-level (and usually proprietary) reactive control. Our tools of choice in architecting robust distributed control systems are the MOOS and/or ROS middleware in conjunction with the excellent interval programming (IvP) Helm developed within the MIT Laboratory for Autonomous Marine Sensing Systems (LAMSS). For high-fidelity simulation of single or multiple interacting autonomous systems, we use the Modular Open Robots Simulation Engine (MORSE), which we have extensively customised to enable detailed simulations of all manner of platforms, sensors and actuators.
Sensing and Perception
Modern approaches to machine learning, including deep learning via convolutional neural networks, are now capable of quickly inferring meaning (or at least labels) to images and sounds in the real world - so called semantic sensing. From searching the depths of the ocean for a missing aircraft to efficiently clearing sea mines in a contested port, or even searching for an injured camper in inaccessible terrain. All of these missions, which would previously have been hamstrung by poor to non-existent communications, suddenly become more efficient and scalable with the advent of even simple artificial intelligence. Mission Systems aims to target those applications and industries which stand to be most radically transformed.
Physics-based simulation means more than just collision detection, and one of Mission Systems' specialties is the accurate simulation of all manner of electro-magnetic, optical and acoustic sensing. Whether your sensor is an electro-optical camera, laser range finder, side-scan sonar or radar, we have the knowledge and tools to engineer realistic synthetic data streams into your systems-level simulations - often in real-time. Central to the fidelity of any sensor simulation is an understanding of the environment, and Mission Systems' expertise in computational ocean acoustics and terrestrial radar propagation allows it to construct detailed forward models, which are an essential component in many detection and imaging applications.
Rapid advances in mobile device technologies are now driving even higher levels of performance in embedded computing. The latest system-on-a-chip solutions for mobile robotics now combine powerful multi-core RISC CPUs with very capable GPUs to realise teraflop performance from minuscule amounts of power. These hardware advances, combined with a renaissance in the theory of artificial neural networks and deep learning, have ushered in a golden age in which mobile robots finally have a useful amount of computational power to deal with complex environments. Mission Systems aims to be at the forefront of these developments with navigation and perception algorithms especially crafted to exploit as much of that power as possible.