Cooperative Control and Planning for Safe Aerial Transportation in Unknown Environments

Abstract

Recently, aerial manipulators using unmanned aerial vehicles (UAVs) are receiving attention due to their superior mobility in three-dimensional space. It can be applied to a wide range of applications such as inspection of hard-to-reach structure or aerial transportation. This dissertation presents a viable approach to safe aerial transportation in unknown environments by using multiple aerial manipulators. Unlike existing approaches for cooperative manipulation based on force decomposition or impedance-based control that often requ- ire heavy or expensive force/torque sensors, this dissertation suggests a method without such sensors, by exploiting the decoupled dynamics to develop estimation and control alg- orithms. With the decoupled dynamics and the assumption of rigid grasp, an online estimator is designed initially to estimate the mass and inertial properties of an unknown payload using the states of the aerial manipulator only. Stable adaptive controller based on the online estimated parameter is then designed using Lyapunov methods. Through simulations, the performance of the proposed controller is compared with conventional passivity-based adaptive algorithms.

This dissertation also proposes a motion generation algorithm for cooperative manipulators to transport a payload safely. If the payload is excessively heavy in comparison with the transportation ability of an aerial robot, an aerial robot may crash because of actuation limits on the motors. As a first step, the allowable flight envelope is analyzed with respect to the position of the end-effector. In order to keep the end-effector in the allowable fight region, kinematic coordination between a payload and cooperative aerial manipulators is first studied. A two-layer framework, in which the first layer computes the motion reference of the end-effectors and the second layer calculates the joint motion of the corresponding manipulator, is then developed in a task-prioritized fashion. When generating aerial manipulator trajectories, the desired trajectory is calculated to satisfy the unilateral constraints obtained by the allowable flight envelope.

This work also considers the obstacle avoidance of cooperative aerial manipulators in unknown environments. Using the relative distance between an aerial robot and an obstacle as measured by an RGB-D camera and point cloud library (PCL), dynamic movement primitives (DMPs) modify the desired trajectory. By having the leader robot detect an obstacle and the follower robots maintain a given relative distance with the leader, improved efficiency of obstacle avoidance for cooperative robots can be achieved.

Finally, the proposed synthesis of estimation, control, and planning algorithms are validated with experiments using custom-made aerial manipulators combined with a two-DOF (Degree Of Freedom) robotic arm. The proposed method is validated with trajectory tracking using two types of payloads. Cooperative aerial transportation in unknown environments is also performed with an unknown obstacle. Both experimental results suggest that the proposed approach can be utilized for safe cooperative aerial transportation.