Over the last 50 years, applications of rotorcraft carrying external suspended loads have been of significant interest in the aerospace research community due to the inherent sta- bility problems such systems suffer from. Rotorcraft suspended load operations have ex- perienced further development and extensive use since the Vietnam war. In the following years (1965-1975) the research turned to the stabilization of difficult loads using heavy lift helicopters (Cicolani et al., 1995). The solutions found included suspensions with multiple attachment points and various control devices. An early successful operation using single helicopter slung load was based on using suspensions consisting of cables and spreader bars (Korsak et al., 1972).
A common obstacle to further operational development is the complexity of the system mo- tion and its guidance and control along manoeuvring flight paths (Cicolani et al., 1986), which somewhat slowed the progress of slung load operations until recently. Progress be- yond hover operations suffered due to the lack of practical and realistic equations of motion for use in simulation studies and, therefore, the development of experimental studies was affected as well.
In the work of (Cicolani et al., 1995) several systematic approaches to derive the equations of motion of the slung load system were identified. Their working equations for applications were formulated almost entirely in terms of the objects and operations of 3-dimensional vector mechanics, their simulation work was demonstrated and now it is used on a number of similar studies.
Another successful example of simulation of single helicopter slung load operation is pre- sented in (Faille et al., 1995). They studied the stabilization and regulation problem of a helicopter/slung-load system by using a non-linear model of the system that was linearised (only a set of equations) in order to apply linear control theory. In their model, the load needs 9DOF (degrees of freedom) that corresponds to 18 states and, when connected to helicopter, the model contains 33DOF resulting in 66 states. An assumption is made that it is possible to measure the position and orientation of the slung load, therefore this method- ology only works in simulation scenarios. Finding information about the current state of the slung load becomes paramount if the control of the load must be addressed. This work
is important for historical reasons, since it proposes a practical methodology to tackle the problem.
For an elaborate survey of the different studies in this area throughout the last decades the reader may refer to the work by (Fusato et al., 2001) and (Luigi, Cicolani, 1992), containing a number of historical references dealing with different types of stability analysis. Two major works are found in (Prasad Sampath, 1980) and (Bisgaard et al., 2006). In the first one, the author used a Lagrange formulation to tackle the modelling problem, being this one of the first studies that created a complete set of equations of motion in 12 degrees of freedom, which included all body-to-body suspension schemes. In the second study, the model was derived using the Udwaidia-Kalaba equation and a redundant coordinate formulation in which the wires were inserted as acceleration constraints.
The majority of studies have been focused on determining stable flight regimes with respect to slung load parameters to avoid instabilities (Prasad Sampath, 1980) (Poli, 1973) (Prab- hakar, 1977). Alternative efforts have considered modifying the shape of the load (Hoh et al., 2006) as well as adding extra components (gyroscopes, fins, drogues) to the load to make it stable related to the rotorcraft (Micale et al., 1973) (Feaster et al., 1977). Adding components of the load is a possible solution, but reduces the applicability and practicability of the overall system, therefore it has to be pondered according to the application.
The problem of state estimation for slung load systems is mentioned sparsely in literature. In (Dukes, 1973) the difficulty in reliably estimate slung load states is mentioned and to overcome this problem an open loop control approach is suggested. (Gupta et al., 1976) considers the design of state estimation for the slung load using an attitude measurement, the angles of a measurement cable from the helicopter to the ground and the angles of the suspension cable as sensor input to a linear Kalman filter.
One particularly relevant study was focused on designing stability augmenting techniques for slung load systems and stability analysis to determine favourable wire lengths, vehi- cle/load mass ratios, and other parameters (Bisgaard et al., 2006). This study resulted in one of the first experimental systems for small scale rotorcraft. Autonomous small scale ro- torcraft have changed the perspective on this field of engineering, making it more accessible for academic research groups and allowing to tackle the problems related with slung load systems. Before this seminal work by Bisgaard, testing of stability augmenting solutions on real aircraft was limited to research and development groups with access to heavy lift, expensive helicopters (military and defence companies).
In more recent work (Bisgaard et al., 2010) presents a design and verification of an esti- mation and control system for a helicopter slung load system. In this work, Bisgaard uses a computer vision approach to create an estimator capable of estimating the states of the load. Vision-based sensor data estimates the relative position of the load and estimates the wire length. After testing the estimator, a feed-forward control system based on input shap- ing is developed so that it enables the helicopter to perform manoeuvres with a slung load without inducing residual oscillations. An estimator for the slung load position inspired by the latter work is created for this thesis.
Tackling the slung load problem using a multi rotorcraft is found in more recent work. (Palunko et al., 2012) addressed the problem of trajectory tracking while carrying a slung
load, using dynamic programming in order to generate a swing-free trajectory for the quadrotor slung load system. The estimation of the slung load states is done by using an indoors motion tracking laboratory.
In (Palunko et al., 2013) a model-free approach to solve the slung load swing trajectory tracking using a reinforcement learning algorithm is proposed. The slung load states are obtained using an indoors motion tracking system. Their method converges quickly to learn the policy function that minimizes the tracking error of the load with respect to the reference trajectory, their results are proved experimentally.
Another approach of simulation work on dynamic modelling of the quadrotor/slung load system can be found on (Sadr et al., 2014), the model was obtained and verified by com- paring two Newton–Euler and Lagrange methods. Their control methodology involved a feed-forward algorithm for reducing or cancelling the swinging load oscillation by imple- menting input shaping theory which convolves the reference command with a sequence of impulses, no experimental work was carry out. A interesting modelling approach for the slung load was presented in (Feng et al., 2015), where they modelled as a three-dimensional point mass pendulum where the dynamics of the slung load are constructed analytically by calculating the suspension angles of the load. An adaptive control scheme is then proposed, it addresses specifically the existence of the external force and torque caused by the slung load. Both modelling and control techniques are verified only with simulations.
A hybrid dynamical system is proposed in (Tang et al., 2015) where the quadrotor is consid- ered as a rigid-body and the load as a point-mass, but their hybrid model comes from two subsystems of models, their numerical and experimental results indicate that the method is practical for generating trajectories that include aggressive obstacle avoidance manoeuvres and hybrid state transitions. The slung load orientation and position is obtained by using an indoors motion tracking laboratory. Attempts using drones in the cargo transport can be
Credit: Raptopoulos, Delft and Momont
Fig. 1.4.: Examples of cargo MRUAV on the healthcare industry.
from from a range of sectors, including the healthcare industry, food, and postal deliveries. In the healthcare industry drones can transport medicines and vaccines, and retrieve medi- cal samples into and out of remote or otherwise inaccessible regions (Andreas Raptopoulos, 2013) (Fig.1.4 bottom left and bottom right). Ambulance drones are developed to rapidly deliver defibrillators in the crucial few minutes after cardiac arrests (Alec Momont, 2014) (Fig.1.4 top right).
Foodwise, there are currently several companies creating delivery services for their food goods. The most relevant are Burrito-by-drone (Fig. 1.5 top left), Domino’s Pizza (Fig.1.5 top right), 7-eleven (Fig.1.5 bottom left) and Old Hamburg Schnitzelhaus AIR (Fig. 1.5 bot- tom right). In the postal delivery sector, postal companies have been forced to seek new ways to expand their traditional letter delivery business models. Different postal companies from Australia, Switzerland, Germany, Singapore and Ukraine have undertaken various UAV trials as they test the feasibility and profitability of unmanned delivery UAV services. Again, the most relevant are Amazon, 2013 (Fig.1.6 top middle), DHL, 2013 (Fig.1.6 bot- tom left) and Google, 2014 (Fig.1.6 bottom right). After reviewing the literature, a very important need for an experimental slung load estimator is found. Such an estimator can then be used on different types of control strategies.