Research
Expanding our knowledge of the ocean with unmanned marine vehicles has long held promise due to the relative safety, higher endurance and lower costs compared to purely human exploration. Unmanned marine vehicles have also been regarded, until recently, as platforms deployed to collect information to be later analyzed by domain experts after the vehicle has been recovered. Due to advancements in the last decade in endurance, on-board computing, through-water and satellite communications, as well the reduction in size and cost of vehicles, an irreversible shift has occurred to expect more on-board decision-making, or autonomy. My work for roughly the last decade and foreseeable future has been focused on research and development of marine autonomy algorithms and architectures to enable unmanned platforms to explore the ocean in ways that are more effective, cheaper, with quicker access to information, and in a way that is more open to contributions from different minds from different fields.
My current research and activities at MIT contains four threads of work that I am pursuing in concert to advance the above overall research objective.
1. Develop new undergraduate and graduate level course material in marine autonomy, sensing and communications. (See MIT 2.680)This is vital to attracting the next generation of young minds interested in exploring the ocean with robotic systems, but also is a terrific environ- ment for better understanding how multiple minds and perspectives may be brought to bear in a group effort to field complex unmanned systems.
2. Development of autonomy algorithms for unmanned surface vehicles in accordance with the Maritime Rules of the Road, or Collision Regulations (COLREGS). Autonomous systems that are capable of safely navigating the ocean with human trust, hold an enormous potential for increasing our ability to sense and understand our environment. The key is to develop methods that are trusted and can be validated to be safe.
3. Better understand what constitutes an effective autonomy system. To this end we will develop unmanned vehicle competitions on the Charles River to gauge relative algorithm effectiveness. Included in this measure will be the issues of algorithm extendibility and simplicity by having an element of the competition metrics remain unknown until shortly before the competition events.
4. Continued development of the MOOS-IvP open source software autonomy architecture. This software is intrinsic to the above three threads and several other MIT funded research projects. It is also used by several other research groups around the world.