Harvest: A Framework and Co-Simulation Environment for Analyzing Unmanned Aerial Vehicle Swarms

Unmanned vehicles have the capability to transform military operations. One relatively unexplored application involves cooperative unmanned vehicle systems called sensor swarms. We propose a conceptual unmanned vehicle swarm: a host of armed reconnaissance vehicles enabling surveillance and targeting (HARVEST). A HARVEST swarm is theoretically capable of autonomous refueling, cooperative search, information fusion, and munitions employment. To enable cooperative swarm capabilities, we identify a set of individual unmanned vehicle services, e.g., localization, querying, and routing. The HARVEST concept, swarm capabilities, and unmanned vehicle services are embodied in our sensor swarm co-simulation environment. The goal is to improve simulation fidelity by integrating existing simulators used within the department of defense (DoD). The process of integrating multiple simulations is known as co-simulation; our design uses OPNET´s external system definition (ESD) to achieve co-simulation. This is the same interface OPNET provides to enable co-simulations based on the high-level architecture (HLA) defined by the defense modeling and simulation office (DMSO). The simulators in the first sensor swarm co-simulation prototype are based on the technology behind NETWARS (OPNET) and the Java programming language. To integrate with OPNET, Java wraps the ESD C-based method calls. This co-simulation is known as a Java, OPNET, and C-Based Co-Simulation (JOCosim). The implementation details of and lessons learned from the first JOCosim prototype are described in this paper. We also briefly discuss the second JOCosim prototype currently under development. The newer version places all simulation control within Java versus just receiving data from OPNET. This capability is crucial to achieving the goal of the second prototype-integrating additional simulation tools such as MATLAB, FalconView, and digital terrain elevation data (DTED)