Mobile Persistent Stare Using Unmanned Systems, 10-R8306
Principal Investigator
Paul A. Avery
Inclusive Dates: 04/02/12 – Current
Background — Within the unmanned ground vehicle (UGV) community there has been a number of successful "short-duration" demonstrations, which perform well for 10 to 45 minutes. In order for UGVs to have long-term success in tactical situations, they will need to perform missions in duration of 4 to 12 hours. The military has indicated that system "resets" could occur during fueling operations that would need to occur at least every 12 hours. The software for these systems needs to be adaptable and flexible enough to allow a system to "correct" itself over the mission duration. Additionally, most demonstrations have used one of two sensing technologies: Light Detection and Ranging (LIDAR) or Electrical Optical (E/O); SwRI is well versed in both technologies, but fusing the two technologies (cooperatively across platforms) could dramatically improve the performance of UGVs. The scope of work in this proposal is to add several additional building blocks to the SwRI MARTI (Mobile Autonomous Research Technology Initiative) platform that are necessary for long-term viability of UGV technology. Groups of unmanned vehicles conducting long-term operations could also benefit from being linked together to form a cooperative vehicle system, which SwRI has experience in implementing for traffic scenarios.
Approach — The approach in this work is to develop hardware-agnostic software tools to automatically collect "health" status data from numerous hardware and software components of an unmanned vehicle, which can later be analyzed for long-term operational effectiveness. This tool will also be integrated into the unmanned vehicle's operating framework, which could enable a more rapid identification and notification of any system warning, error, or failure. With this tool, we will be able to conduct endurance autonomy testing with SwRI's fully autonomous HMMWV 1165. A perimeter will be identified for a surveillance mission around the SwRI facility. A cooperative perception system framework will also be developed to allow the integration of sensor and position data from a number of UGVs. We will define and develop tactical behaviors for the cooperative system and integrate these into the assessment of a cooperative vehicle mission.
Accomplishments — Hardware-agnostic software tools have been developed using the Robot Operating System (ROS) diagnostic tools for logging data related to the operation and behavior states of an autonomous vehicle platform. The specific platform that has been targeted initially is the SwRI-owned HMMWV 1165. The data logging system functions by subscribing to ROS topics that indicate the status of various subsystems, such as behavior states and health status of all included sensors. This manner of operation allows for the system to operate without any modifications to the current vehicle codebase. During operation, the data is periodically saved to a human-readable file called a YAML file, which is later moved to another computer for processing. The data can be parsed in a variety of ways, and presented visually, to further facilitate analysis of the performance of the vehicle during the test. A model of a cooperative vehicle system has been developed using agent-based methods, where a simplified UGV model (agent) is combined with other agents in a scenario of perimeter patrol. The behavioral characteristics of individual UGVs when some specified anomaly is either detected or communicated from another vehicle are being investigated regarding their effect on system-level behavior.
