Guidance and Path-Planning Systems for Autonomous Vehicles
Abstract
This thesis is concerned with two interconnected and very important problems regarding the autonomy of vehicles, namely, path planning and guidance. By adopting a modular approach, path planning and guidance can be viewed as two modules which belong to a wider context consisting of four modules, the other two being navigation and control. All four modules interact with each other and none is completely independent. Path planning deals with what we want to achieve (by defining spatial and temporal constraints), and guidance dictates how we should act in order to achieve it (by generating appropriate reference trajectories to be fed to the corresponding controllers). Therefore it is important to develop: a) path design methodologies, which will generate feasible and safe paths with several desired properties, and b) guidance laws capable of generating reference trajectories which will lead the vehicle on the desired path, even when unknown disturbances (such as ocean currents and wind forces) affect the vehicle’s motion.
Four contributions pertaining to the path-planning problem are included in this thesis. The most important is the use of Fermat’s spiral (FS) as an alternative to both Dubins paths and clothoids. We show that paths consisting of straight lines and FS arcs are curvature-continuous, computationally inexpensive and can be used for path tracking by changing the parametrization. The second contribution is the development of a number of path-evaluation criteria which aim at providing an onboard computer with sufficient information for selecting the right path for a given application. The methodology is still at its infancy but several improvements, which could result in fast progress, are discussed. The third contribution is the use of a monotone cubic Hermite spline for path-planning purposes. The main advantage is that the method generates very practical paths which do not include wiggles and zig zags between two successive waypoints. Moreover, the method provides the user with better shape control, a property which can turn out to be very useful in real-time collision-avoidance applications. The fourth contribution pertains to a collision-avoidance strategy combining the Voronoi diagrams (VD) method and FS-based path generation. An intuitive and efficient procedure is developed for obtaining smooth paths which keep the vehicle at a safe distance from all obstacles on the map and at the same time avoid unnecessary heading changes.
The thesis also presents a number of guidance-related contributions, each of varying degree of importance and difficulty. The first one is the modification of the line-of-sight (LOS) guidance by introducing a time-varying equation for the lookahead distance . This aims at obtaining a more flexible behavior regarding the steering of the vehicle because for very small the vehicle approaches the target path at a direction almost normal to the path, whereas for very large it takes a longer time for the vehicle to converge to the path. The effect of the time-varying equation from a stability viewpoint is investigated. The second (and minor) contribution is the consideration of a 5-DOF vehicle kinematic model (common for torpedo-shaped underwater vehicles which do not control the roll angle) and the influence of the coupling between the horizontal and vertical planes on the expression for the sideslip angle. This led to the third contribution, which is a transformation of the LOS guidance in quaternion form for both the uncoupled and the coupled cases. The transformation is based on exploiting very simple trigonometric properties and the geometry of the LOS guidance. The fourth contribution is an integral LOS guidance law capable of eliminating the errors induced by constant external disturbances. The method is formulated using absolute velocity-based vehicle kinematics and simple Lyapunov-based analysis. The fifth contribution moves a few steps further and presents two adaptive integral LOS guidance laws which compensate for the errors induce by ocean currents. These methods are based on the vehicle kinematics in relative-velocity form. This is a very useful result for underwater vehicles, where absolute velocity measurements might not be available. The effect of the current on the direction normal to the direction of motion (that is, the force inducing the cross-track error) is estimated, and stability results for curved paths are also given. The sixth contribution is the development of a guidance technique where, in addition to the LOS guidance for minimizing the cross-track error, surge velocity commands are generated as well in order to minimize the along-track error, hence satisfying constraints related to the path-tracking (or trajectory-tracking) motion control scenario. Finally, the path-tracking solution is combined with the indirect adaptive integral LOS so as to achieve path tracking under the influence of ocean currents, which also results in estimating all the parameters of the current (that is, current velocity and orientation w.r.t. the inertial frame).
In all cases, particular emphasis was placed on finding solutions that are simple and, at the same time, efficient.