Real-Time Optimal Guidance and Obstacle Avoidance for UMVs

Abstract

The single most important near-term technical challenge of developing an autonomous capability for unmanned vehicles is to assess and respond appropriately to near-field objects in the path of travel. For unmanned aerial vehicles (UAVs), that near field may extend to several nautical miles in all directions, whereas for unmanned ground and maritime vehicles, the near field may only encompass a few dozen yards directly ahead of the vehicle. Nevertheless, when developing obstacle avoidance (OA) manoeuvres it is often necessary to implement a degree of deliberative planning beyond simply altering the vehicle’s trajectory in a reactive fashion. For unmanned maritime vehicles (UMVs) the ability to generate near-optimal OA trajectories in real time is especially important when conducting sidescan sonar surveys in cluttered environments (e.g., a kelp forest or coral reef), operations in restricted waterways (e.g., rivers or harbours), or performing featurebased, terrain-relative navigation, to name a few. For example, a primary objective of sidescan sonar surveys is 100% area coverage while avoiding damage to the survey vehicle. Ideally, a real-time trajectory generator should minimize deviations from the preplanned survey geometry yet also allow the vehicle to retarget areas missed due to previous OA manoeuvres. Similarly, for operations in restricted waterways, effective OA trajectories should incorporate all known information about the environment including terrain, bathymetry, water currents, etc. In the general case, this OA capability should be incorporated into an onboard planner or trajectory generator computing optimal (or near-optimal) feasible trajectories faster than in real time. For unmanned undersea vehicles (UUVs) the planner should be capable of generating full, three-dimensional (3D) trajectories, however some applications may require limiting the planner’s output to two-dimensions (2D) for vertical-plane or horizontal-plane operating modes. For unmanned surface vehicles (USVs) the latter case is the only mode of operations. Consider a typical hardware setup consisting of a UUV augmented with an autopilot (Fig.1). The autopilot not only stabilizes the overall system, but also enables vehicle control at a higher hierarchical level than simply changing a throttle setting ( ) T t δ , or deflecting stern plane ( ) s t δ or rudder ( ) r t δ angles. In Fig.1, WP x , WP y , WP z are the vectors defining x, y, and z coordinates of some points in the local tangent (North-East-Down (NED)) plane for waypoint navigation. Alternatively a