Abstract
Algorithms for autonomous watercraft in swarms or standalone configurations are developed and presented in this work. Motivation of this work is the increase of safety at sea and the improved efficiency of waterborne transport. The purpose of the research is to contribute toward systems of standalone and swarms of surface watercraft similar to those for unmanned aerial vehicles. Rotorcraft as well as fixed-wing aircraft rely on a mix of control theoretical approaches with techniques of machine learning and artificial intelligence. The contribution of the work toward advancing the science and engineering of swarm watercraft autonomy. The methodology used is mainly computational with substantial analytical support of the results. Approaches investigated toward the development of autonomous watercraft and swarms involve distributed, i.e. not centralized, control and intelligence. It has been argued that decentralization of the control structure and task can enhance robustness of the controllers involved and improve scalability. The concept is adapted and applied to small, low-cost surface watercraft. However, in the future it can be expanded to include underwater, aerial and even amphibious vehicles. Various scenarios are investigated. All watercrafts are assumed to have identical physical characteristics and dynamic response but with small fluctuations and perturbations superimposed to imitate dissimilarities due to chance. The surface watercraft dynamics are simplified to model forward motion and rotation about yaw axis; more detailed and precise models can also be used if desired or required. The local-loop controllers, which conventionally are called autopilots, ensure that the acceleration of the craft is adhering to the one dictated by the higher-level supervisory control system. Synthesis of the control law involves amongst other feasibility of the derived governing dynamics as well as energy consumption, time to terminal conditions etc. Computationally efficient algorithms are examined for optimal route planning and obstacle or collision avoidance, by porting the significant progress that has been performed in recent years for terrestrial and space exploration robots. The problem of navigation and route planning of an autonomous craft is tackled by a variety of algorithms. A very important aspect of any algorithms that will be taken into account is computational complexity in conjunction with efficiency. Efficiency can be roughly defined as the inverse of complexity for any algorithm. The practical result of computational efficiency obviously lies in the requirements from the embedded system that forms the target hardware. This is why a real-time implementation study, including a complexity analysis, is intended for all algorithms developed. For both the control and navigation algorithms, a distributed form has been sought after, so they can be implemented on low-cost digital electronic hardware onboard the watercrafts themselves. The hardware developed consists of a low-end microcontroller, an assisting CPLD and the peripherals that perform data acquisition and drive the actuators. After the onboard module architecture was defined, the control algorithms are ported to the target hardware and tested for functionality. Such a distributed approach offers except lower cost, increased redundancy and flexibility compared to the fully centralized approach.
