Event Location: Newell Simon Hall 1109

Abstract: Currently deployed unmanned aerial vehicles (UAVs) rely on preplanned missions or teleoperation and do not actively incorporate information about obstacles, landing sites, wind, position uncertainty, and other aerial vehicles during online trajectory planning. However, to enable autonomous missions in cluttered environments it is necessary to react to all available information during mission execution.

Prior work has successfully addressed sub-problems such as avoiding obstacles, searching for targets, or landing at a known location. In this thesis, we develop a framework for planning that considers flexible mission goals such as searching for and landing at a previously unknown landing site. This problem differs from current planning approaches where typically a goal coordinate is specified.

In the framework, several sub-objective functions, constraints, and a state machine define the mission and behavior of an UAV. As the vehicle gathers information by moving through the environment, the objective functions account for this new information. The user in this framework specifies a function that defines the mission rather than giving a sequence of prespecified goal points. The objective is used in a trajectory optimization algorithm to determine the best path the robot should take.

In initial work, we examine the mission of finding a place to land near a human that needs to be picked up. Good landing sites and obstacles are unknown, and the wind direction can change. We compare a simple algorithm with a fixed plan, to an algorithm that plans using the feedback from obstacles and a map of good places to search. Incorporating the feedback results in a speedup of 3.5-6.2 times in mission completion time and enables operation near obstacles. The landing site evaluation algorithm uses a patch-based coarse evaluation for slope and roughness, and a fine evaluation that fits a 3D model of the helicopter and landing gear to calculate a goodness measure. Results of the algorithm tested are discussed for ladar data collected on a real helicopter.

We propose several improvements to the simple trajectory optimization algorithm in future work. First, a principled method to combine several objective functions is necessary. Second, the trajectory optimization needs to be combined with a planning algorithm to cover a larger set of hypothesis. Third, the constraints of the objective functions, such as dynamics and obstacles, need to be projected to ensure that the optimized trajectory is feasible. Additionally, we propose to improve the state switching to incorporate the sensing uncertainty since the measurements of switching events are uncertain. In future experiments, we intend to test the algorithms on an autonomous helicopter. The framework and algorithms developed in this thesis are going to enable safe and efficient operation of unmanned aerial vehicles in changing and uncertain terrain.

Committee:Sanjiv Singh, Chair

Takeo Kanade

Alonzo Kelly

Emilio Frazzoli, Massachusetts Institute of Technology