Decentralized UAV guidance using modified boid algorithms
Decentralized guidance of Unoccupied Air Vehicles (UAVs) is a very challenging problem. Such technology can lead to improved safety, reduced cost, and improved mission efficiency. Only a few ideas for achieving decentralized guidance exist, the most effective being the boid algorithm. Boid algorithms are rule-based guidance methods derived from observations of animal swarms. In this paper, boid rules are used to autonomously control a group of UAVs in high-level transit simulations. This paper differs from previous work in that, as an alternative to using exponentially scaled behavior weightings, the weightings are computed off-line and scheduled according to a contingency management system. The motivation for this technique is to reduce the amount of on-line computation required by the flight system. Many modifications to the basic boid algorithm are required in order to achieve a flightworthy design. These modifications include the ability to define flight areas, limit turning maneuvers in accordance with the aircraft dynamics, and produce intelligent waypoint paths. The use of a contingency management system is also a major modification to the boid algorithm. A Simple Genetic Algorithm is used to partially optimize the behavior weightings of the boid algorithm. While a full optimization of all contingencies is not performed due to computation requirements, the framework for such a process is developed. Wolfram's Matlab software is used to develop and simulate the boid guidance algorithm. The algorithm is interfaced with Cloud Cap Technology's Piccolo autopilot system for Hardware-in-the-Loop simulations. These high-fidelity simulations prove this technology is both feasible and practical. They also prove the boid guidance system developed herein is suitable for comprehensive flight testing.