Aerial robot navigation in cluttered urban environments

Autonomous navigation systems for mobile robots have been successfully deployed for a wide range of planar ground-based tasks. However, very few counterparts of the previous planar navigation systems were developed for three-dimensional (3-D) motion, which is needed for unmanned aerial vehicles (UAVs). Safe maneuvering in complex environments is a major challenge for UAVs. Future urban reconnaissance and search missions will require UAVs to autonomously navigate through cluttered urban spaces. This research proposes two approaches for unmanned helicopter navigation in cluttered urban environments: a 3-D fuzzy behavioral approach and a 3-D vector field histogram (VFH) approach.; Behavior-based control has been very successful for planar mobile robots navigation in unknown environments. A novel fuzzy behavioral scheme for navigating an unmanned helicopter in cluttered 3-D spaces is developed. The 3-D navigation problem is decomposed into several identical two-dimensional (2-D) navigation sub-problems, each of which is solved by using preference-based fuzzy behaviors. Due to the shortcomings of vector summation during the fusion of the 2-D sub-problems, instead of directly outputting steering subdirections by their own defuzzification processes, the undefuzzified intermediate results of the sub-problems are fused to a 3-D solution region, representing degrees of preference for the robot movement. A new defuzzification algorithm that steers the robot by finding the centroid of a 3-D convex region of maximum volume in the 3-D solution region is developed. A fuzzy speed control system is also developed to ensure the efficiency and safety of the navigation.; The VFH approach is very popular for planar mobile robots. A 3-D VFH approach to UAV navigation in cluttered urban environments is developed. A 3-D laser measurement system is used to obtain the obstacle distribution in this method. Instead of a 2-D Cartesian histogram grid as a world model, a 3-D spherical histogram mesh is applied. This 3-D histogram mesh is updated continuously with range data. The 3-D VFH method subsequently employs a two-stage data-reduction process in order to compute the desired control commands for the robot. In the first stage the 3-D histogram mesh is reduced to a 2-D polar histogram corresponding to all possible steering directions for the robot. In the second stage, a novel convex finding algorithm is applied to efficiently find candidate directions from the 2-D polar histogram. The most suitable sector within the candidates with the lowest value of a particular cost function is selected, and the steering of the robot is aligned with that direction.; Substantial simulations have been carried out to demonstrate that the two algorithms proposed in this dissertation can smoothly and effectively guide an unmanned helicopter through unknown and cluttered urban environments. Comparison simulation results show that the 3-D VFH has the ability to travel shorter and smoother pathes at most of scenarios. However, the feature doesn't apply to the 2-D counterparts. The 2-D fuzzy behavioral method usually has a smoother path, but the 2-D VFH travels a shorter path in most of scenarios.