Testing multi-agent autonomy in complex communications environments

摘要

Advanced approaches to intelligent, real-time decision making are difficult to test for multi-agent systems. Specifically, the expense and logistics of fielding multiple unmanned underwater vehicles in an open ocean environment are significant challenges in marine autonomy. Traditionally, algorithms may be developed and tested under ideal conditions with tools that do not accurately reflect real-time performance (e:g: MATLAB). Effort is then required to integrate these algorithms onto the hardware platforms intended for in-water experimentation, so that they may be validated in a realistic setting. A simulation platform that enables the execution of complex algorithms using realistic sensor models while minimizing integration efforts enhances the ability to more productively prototype autonomy. This work highlights developments for a real-time autonomy simulator based on the common marine autonomy architecture, MOOS (Mission-Oriented Operating Suite), and DRI (Distributed Resource Integration), named DRI-MOOS. This unique pairing of software solutions minimizes that hardware integration requirement and expands the potential use cases for testing real-time vehicle performance and in situ decision making. This work presents example applications that use this simulator to demonstrate the impact of real-time sensor simulation for such a purpose.