Prev-Talk | Next-Talk | All-Talks | Talks-Sorted

Talk-02: Real-time Simulation of Side Scan Sonar within the MOOS framework using NVIDIA OptiX

Trevor G. Anderson, Australian Centre for Field Robotics, The University of Sydney

David G. Johnson, Australian Centre for Field Robotics, The University of Sydney

David Battle, Defence Science and Technology Organisation

Autonomous underwater vehicles (AUVs) are regularly used as platforms for side scan sonar systems. One particularly common use is in the detection, identification and classification of underwater mines; known as mine countermeasures (MCM). Traditionally the development of processing algorithms for this task is performed iteratively; that is, a vehicle is taken out to an example mine field, either real or populated with training mines, where a selection of scans are collected. The vehicle simply records the collected data including navigational information which is then processed back at the “office” to train and improve the algorithms. When as much as possible has been gained from the latest set of data another trial is run, hopefully under different conditions so as to be useful in training.

Since changes in the ocean floor are limited over reasonable time scales and continually laying new realistic synthetic mine fields is difficult groups are forced to take their vehicles further and further on field trips at great expense; both financially and in lost development time. In collaboration with the Australian Defence Science and Technology Organisation (DSTO) we have developed a high fidelity, highly configurable real time side scan sonar simulator such that virtual trials in variable mine fields can facilitate rapid iterative development.

Our simulator integrates with a MOOS simulation enabling complete missions to be run including all autonomy systems, acoustic communication and vehicle dynamics and control. The MOOS simulator provides regular updates on the vehicles real world position as well as its internal position estimate, which our simulator uses in combination with a high fidelity 3D mesh of the environment to provide simulated side scan sonar output, including estimated position information for each pulse. The side scan simulator itself uses the NVIDIA OptiX ray tracing framework to perform the core of the simulation on any available CUDA enabled compute card(s). In this presentation we present the simulator as applied particularly to a REMUS 100 vehicle, as that is the primary vehicle in use by DSTO for side scan based MCM. DOWNLOAD

Categories:

  • AUVs
  • MCM
  • Simulation/Visualization