This article presents a new algorithm for short-term maritime collision avoidance (COLAV) named the branching-course model predictive control (BC-MPC) algorithm. The algorithm is designed to be robust with respect to noise on obstacle estimates, which is a significant source of disturbance when using exteroceptive sensors such as, for example, radars for obstacle detection and tracking. Exteroceptive sensors do not require vesselto-vessel communication, which enables COLAV toward vessels not equipped with, for example, automatic identification system transponders, in addition to increasing the robustness with respect to faulty information which may be provided by other vessels. The BC-MPC algorithm is compliant with Rules 8, 13, and 17 of the International Regulations for Preventing Collisions at Sea (COLREGs), and favors maneuvers following Rules 14 and 15. Specifically, the algorithm can ignore the specific maneuvering regulations of Rules 14 and 15, which may be required in situations where Rule 17 revokes a stand-on obligation. The algorithm is experimentally validated in several fullscale experiments in the Trondheimsfjord in 2017 using a radar-based system for obstacle detection and tracking. To complement the experimental results, we present simulations where the BC-MPC algorithm is tested in more complex scenarios involving multiple obstacles and several simultaneously active COLREGs rules. The COLAV experiments and simulations show good performance. K E Y W O R D S control, marine robotics, planning

This paper addresses the challenges of making safe and predictable collision avoidance decisions considering uncertainties related to maritime radar tracking. When a maritime radar is used for autonomous collision avoidance, strategies for handling uncertain obstacle tracks, false tracks, and track loss become necessary. Robust decisions are needed in order to achieve clear and predictable actions according to the international regulations for preventing collisions at sea (COLREGs). We present robustness considerations and results of using an Integrated Probabilistic Data Association (IPDA) tracking method with a collision avoidance method based on Model Predictive Control. The results are from full-scale experiments that cover challenging multiple dynamic obstacle scenarios, including realistic vessel interactions where some obstacles obey COLREGs, while others do not.

Collision avoidance systems are a key ingredient in developing autonomous surface vehicles (ASVs). Such systems require real-time information about the environment, which can be obtained from transponder-based systems or exteroceptive sensors located on the ASV. In this paper, we present a closedloop collision avoidance (COLAV) system using a maritime radar for detecting target ships, implemented on a 26 foot highspeed ASV. The system was validated in full-scale experiments in Trondheimsfjorden, Norway, in May 2017. The probabilistic data association filter (PDAF) is used for tracking target vessels. The output from the PDAF is processed through a least-squares retrodiction procedure in order to provide the COLAV system with sufficiently accurate course estimates. A tracking interface provides estimates of target states to the COLAV system, which is based on the dynamic window (DW) algorithm. DW is a reactive COLAV algorithm originally designed for ground vehicles, and we therefore make a number of modifications to adapt it for use with high-speed ASVs. The closed-loop experiments demonstrated successful COLAV with this system, but also disclosed several challenges arising from both the DW algorithm and the tracking system, motivating for further work.

No abstract

No abstract