A heuristically constrained dynamic perception architecture for the Explosive Ordnance Disposal Autonomous Underwater Vehicle Robotic Work Packages Program

摘要

A unique mission control and vehicle management architecture which facilitates machine-based interaction with underwater ordnance has been derived for the Explosive Ordnance Disposal Autonomous Underwater Vehicle Robotic Work Packages Program (EODRWP). A number of key concepts have been integrated to provide for deliberative sequential and reactive mission plan execution. These include dynamic perception, which provides for active sensor management and vehicle position planning for enhanced classification, and heuristic constraint which incorporates knowledge of target and vehicle characteristics to provide a basis for interaction with the object of interest. This paper discusses the architectural framework which was developed in the first year of the program by enhancing a real-time expert system to implement a combined mission planning and constraint executor. Integration with a subsumptive layer which implements behaviors through a directed-task approach is described as are interactions with a vehicle-specific controller and sensors.