Ferries are irreplaceable elements of public transport systems in many countries. Whenever water needs to be passed, and a bridge is not feasible, there are no alternatives. Ferries often operate on relatively short legs but amongst uncontrolled environments and other water traffic, meaning the task might be repetitive and boring for a human pilot but nontrivial for a computer. To let skilled pilots focus on the challenging routes, Reaktor AI has built an autonomous ferry concept to save operating costs on the boring ones by having one person remotely monitor multiple ferries.
Despite the developments in self driving cars, the autonomous operation of freely moving vehicles is never easy. Let alone in uncontrolled or loosely marked environments that are without, for example, air traffic control or clear traffic signs. Although the task of driving a ferry from a pier to the one on the opposite side of a channel and docking to it is entirely feasible, there is a lot more to it. The environment is complex and even small ferries are slow to change course or to stop. How can you ensure a ferry does not sail over swimmers? How can you avoid crossing traffic or leisure boats that don’t respect traffic rules? How can you call for help if a passenger goes overboard?
The small-scale ferry concept made by Reaktor AI has been built to serve as a platform for research and development instead of being implementation ready per se. Currently, it operates between two marked piers using, for example, real-time neural networks to navigate and avoid obstacles, and it offers a good sandbox, for instance, for sensor fusion development.