Providing Nautical Chart Awareness to Autonomous Surface Vessel Operations
|Title||Providing Nautical Chart Awareness to Autonomous Surface Vessel Operations|
|Publication Type||Conference Proceedings|
|Conference Name||OCEANS 2016 MTS/IEEE Monterey|
|Conference Dates||September 19-23|
|Conference Location||Monterey, CA|
|Keywords||asv, ENCs, MOOS-IvP, Obstacle Avoidance|
When a mariner navigates into an unfamiliar area, he/she uses a nautical chart to familiarize him/herself with the environment, determine the locations of hazards, and decide upon a safe course of travel. An autonomous surface vehicle (ASV) would gain a great advantage if, like its human counterpart, it can learn to read and use the information from a nautical chart. Electronic Nautical Charts (ENCs) contain extensive information on an area, providing indications of rocks and other obstructions, navigational aids, water depths, and shore lines. The goal of this research is to increase an ASV’s autonomy by using ENCs to provide guidance to the helm when its intended path, which may be dynamically changing, is unsafe due to known hazards to navigation, and context to its sensor measurements that are invariably subject to uncertainty.
The approach taken in this paper divides nautical chart awareness into two sections: obstacle avoidance and contextualizing sensor measurements. Unplanned changes to the ASV’s path, such as avoidance of other vessels or previously unknown obstacles sensed by the ASV in real-time, may cause the ASV to maneuver into an unsafe environment. Prior mission planning, even with knowledge of nautical charts, cannot account for these dynamic responses. Therefore, to navigate an ASV safely through its environment, obstacle avoidance procedures have been developed to reactively change the ASV’s path to avoid known obstacles identified from ENCs. The ENC obstacle avoidance procedures are implemented in a behavior-based architecture where information on the potential threat of the nearby obstacles, as well as the ASV’s current state, are used to penalize heading choices that would intersect with the obstacle and, when combined with the waypoint behavior, ensures safe travel around the obstacle while maintaining close proximity to the original path.
Identifying objects in a camera, sonar, LIDAR or other sensor’s data can be a challenging endeavor in an ocean environment due to the variable sea state, wind, fog, sea spray, sun glint from the sea surface, and bubbles in the water column. Therefore, providing a prior probability distribution for the likely location of those objects in a sensor’s field of view has the potential to significantly enhance object detection processing. Contextualizing sensor measurements dynamically identifies objects from the ENC in a sensor’s field of view and provides that information to the sensor in real-time.
To accomplish these tasks, feature layers within a standard ENC must be translated to a spatial database. In this database, features are encoded with a “threat level” based on the feature type and the estimated depth of the object, which is not always encoded within the ENC. Variations in the local tides as well as the vessel size and speed are also factors when deciding the threat level and the vehicle’s appropriate course of action.
Providing an ASV the ability to read, understand, and use nautical charts allows the ASV to safely react to known obstacles in its environment and to increase robustness of sensor detection algorithms. No mariner would go into an unfamiliar harbor with restricted visibility without consulting a nautical chart. Autonomous surface vehicles should not be an exception.
|URL for Proceeding||http://ieeexplore.ieee.org/abstract/document/7761472/|