Evaluation of an advanced slung load control system for piloted cargo operations

Helicopter operations with externally slung loads are highly demanding for the flight crew. As the pilot does not have a direct view of the load, assistance is required from an additional crew member to achieve operational requirements (e.g., for load handling to achieve precise load positioning). An automatic load stabilization and positioning system for cargo operations has been designed with the aim to reduce pilot workload, damp load pendulum motion and to improve the load positioning performance. This system uses the concept of load-motion feedback to the rotor control. To avoid degradation of Handling Qualities (HQs), as found in previous investigations, a function has been developed that monitors pilot control inputs. Dependent on the amplitude and duration of pilot control stick deflection, the feedback signal for slung load damping is blended between two different gain sets. One set provides improved HQs during piloted control and one set provides good load damping when the pilot is passive. A further novel aspect is the evaluation of an automatic load control system using a Translational Rate Command as a method of helicopter control. A piloted simulation study has been conducted using this advanced load control system with automatic load stabilization and positioning. Three test pilots evaluated the system in different control law configurations using a Mission Task Element simulating an external load cargo operation. HQs and pilot workload were evaluated using the Cooper-Harper Rating Scale and NASA Task Load Index, respectively. The results of the study show that improved HQs and reduced pilot workload in combination with improved task performance can be achieved with the advanced slung load control system.


Introduction
Helicopter operations with an external load suspended by a sling are considered high-risk operations as they hold many potential dangers (e.g., mechanical failure, load clearance, reduced flight performance, degraded Handling Qualities (HQs)) and, therefore, are highly demanding for the whole crew [1]. In general, the pilot has no direct view on the load and has to rely on verbal instructions of crew members or on devices such as a mirror. Particularly, during the load pickup and set-down, the pilot is exposed to a high workload, since the pilot has to stabilize the helicopter in hover and control the load motion simultaneously. It is hard to suppress load swing and at the same time place the load precisely on a target position.

Slung load control systems
In the past, numerous concepts and systems for improving the control of the lightly damped pendulum motion of an external load have been investigated and tested in flight [2][3][4][5][6]. According to Ivler et al. [4], the systems can be classified into two main categories: the direct (or on load control mechanism) and the indirect control mechanism. The direct control mechanism generates control forces or moments directly on the slung load to increase effective load damping (e.g., an active load hook that can be moved relative to the helicopter to damp the load swing). This is independent from the motion of the helicopter fuselage. The indirect approach controls the load through displacements and rotations of the entire helicopter. To achieve this, the load motion is fed back to the rotor control channels. Due to its low system complexity and weight, the concept of feeding back the load motion to the rotor controls was used for external load control in the latest study.
Several works have investigated the impact of the load dynamics on the piloted handling of the helicopter with a suspended load and how to improve both the HQs and load damping [4][5][6]. Ivler et al. [4] found that a fundamental trade-off between HQs and load damping exists for indirect control mechanism. Good HQs with an external load can only be achieved at the expense of degraded load damping and vice versa.
Within the Heavy-Lift Helicopter program in the 1970s, a first system for load positioning was developed for the flight demonstrator Boeing Model 347 [7]. The demonstrator was equipped with a retractable cabin for the loadmaster who was sitting rearwards and facing the load. Over an additional control stick, the loadmaster was able to maneuver the helicopter with low control authority. With engaged load stabilization, a precise load positioning could be demonstrated in flight test. In the 1970s the electronic hardware was large, heavy and expensive thus a further development of this technology was stopped. In the early 2000s, the idea of automatic slung load control was rediscovered. Improvements over the last decades in electronic hardware now allows for comparatively easy system integration with reduced weight. The first successful demonstration in the field of unmanned full-size helicopters was reached with a highly modified K-MAX helicopter called BURRO (Broadarea Unmanned Responsive Resupply Operations). This helicopter is able to deliver several cargo loads at different locations even autonomously by flying a pre-programmed course [8]. Only few details about the slung load control are available in the literature [3]. A combination of indirect load stabilization by cable angle feedback and a special design of the cargo hook trolley assembly [9] were used to improve load handling.
Recent works in the USA have investigated the use of a combination of direct and indirect load control mechanism realized through an active cargo hook and load motion feedback to the rotor controls (Refs. [5,10,11]). The active hook adds an additional degree of freedom to the system. The hook can be automatically positioned relative to the helicopter to damp the load swing to some extent, in addition to the motion of the helicopter body. Using both control concepts allows HQ and load damping requirements to be met simultaneously. The drawbacks of such hybrid systems are increased take-off-weight and additional system complexity.
The German Aerospace Center (DLR) recently completed the project SISAL (Sicherheitsrelevante Systeme und Ansätze in der Luftfahrt, english: safety-relevant systems and approaches in aviation) ( [6,12,13]) in collaboration with Airbus Helicopters Deutschland GmbH and iMAR Navigation GmbH. The primary objective of SISAL was the design and evaluation of an Automatic Load Control System (ALCS) based on load motion feedback to the rotor controls for both forms of load suspension: loads suspended from a rescue hoist and a cargo hook. Two functions were developed during the project, the Automatic Load Damping System (ALDS) and Automatic Load Positioning System (ALPS).
During the first flight tests using the ALDS for hoist operations with DLR's research helicopter, ACT/FHS (Active Control Technology/Flying Helicopter Simulator) [6] (see Fig. 1), a degradation in pilot handling and load damping was apparent when the pilot was controlling the helicopter with active load control in hover. This control conflict was investigated in a subsequent simulator study [12] with the focus of hoist operations in hover. Without manual pilot control (i.e., with the pilot 'hands off'), the load damping was successfully demonstrated during the flight tests [6].
When the automatic load stabilization was active and the pilot was simultaneously controlling the helicopter by control inputs in hover condition the load oscillations became unstable (see Fig. 2). This was due to the requirement of the load damping controller to move the helicopter to damp the load motion. At the same time the pilot is attempting to hold the position of the helicopter and, therefore, acting against the load damping controller commands (see Fig. 2). As a result, the helicopter response to the stick input becomes unpredictable, resulting in degraded HQs. In Fig. 3, a case is shown, where the pilot went 'hands off' after an initial longitudinal stick input to excite a load swing. After about 15 s, ALDS is switched on and the load is automatically stabilized.
In this study it was concluded that one solution to resolve the inherent conflict between pilot control and load control in hover is to remove the pilot from the control loop during the load positioning phase. To accomplish this, the ALPS was developed [13]. The ALPS can be activated by the pilot in hover and the load can be automatically positioned over a commanded target position. For low speed and forward flight, when the pilot is manually controlling the helicopter by his stick inputs, the ALDS provides additional automatic load damping. Both systems combined are the overall ALCS.

New research
In 2017, first flight tests were conducted using the ALCS on an Airbus Helicopters H135 prototype machine for a centrally mounted cargo hook configuration. The upper part of Fig. 4 shows the hook beam with an integrated slung load sensor, developed by iMAR Navigation GmbH, and the rope marker which are used to measure the slung load motion. In the lower part of Fig. 4 the cockpit of the H135 helicopter   Due to the limited time for testing, the flight tests with the H135 helicopter and ALCS were only the proof of concept for the sensor concept and closed-loop performance of the load control system in a near-serial testbed. Therefore, a comprehensive piloted simulator study using DLR's Air Vehicle Simulator (AVES) was conducted to investigate the performance, benefits and deficiencies of a load control system for piloted operations in depth. The simulator study allowed to test different control systems easier and cheaper and to evaluate the system by several test pilots.

Paper objectives
In this paper, an advanced ALCS is presented and evaluated in a piloted simulation study with a simulation model of the ACT/FHS, with a centrally mounted cargo hook.
All the data presented in this paper has been obtained using a simulation model based on the ACT/FHS. The ACT/ FHS is a highly modified EC135 with a fly-by-wire/fly-bylight control system [14]. For this reason, the data presented here, including the vehicle responses and the predicted and assigned Handling Qualities Ratings (HQRs) are not directly comparable to any helicopter from serial production.
Two new aspects regarding the ALCS are presented in this work. First, a load control design that adjusts the load feedback to improve the HQs when automatic load stabilization is active and the pilot is manually controlling the helicopter. Second, the ALCS is tested in combination with a Translational Rate Command (TRC) mode as basic helicopter control law. The TRC is an upper control law mode which is able to provide very high stability and easy handling of the helicopter in low-speed and hover condition. The TRC used here was rated in Handling Qualities Level 1 for the helicopter without slung load [15].
The focus of this paper is the design consideration and optimization of the ALCS, and the evaluation of the advanced ALCS in a piloted simulation study. The paper continues as follows: First, control law structures are introduced and the optimization strategy of the ALCS is presented in detail. Afterwards the test set-up for the piloted simulation is explained and the evaluation results are comprehensively discussed. This includes the data analysis of the results in terms of HQRs, workload ratings and quantitative performance data. At the end of the paper conclusions are drawn.

Control system design
For the ALCS design, a comprehensive tool chain using CONDUIT was built for the application with DLR's flying testbed ACT/FHS and simulator AVES as described in [13].

Automatic load control
The load control system features the load stabilization system and the load positioning system. The aim of the load stabilization is to increase the damping of the load pendulum motion by reducing the load motion in relation to the helicopter which is described by the angles ( c , c ) between the load suspension (cable) and the inertial vertical helicopter axis in Fig. 5.
During piloted control with automatic load stabilization active, the helicopter motion from the load control can significantly differ from the motion commanded by the pilot which can result in degraded HQs. As suggested by [16], one possible approach is a control system that provides a control mode for piloted and automatic handling. The idea is that the mode for pilot handling is especially optimized to provide good HQs and acceptable load damping. And the mode for automatic handling is optimized for high load damping and can be used when the pilot is out of the control loop. This idea has been adopted and applied to the stabilization system presented in this paper.
For precise load handling in hover, an automatic function for load positioning is provided. The aim of load positioning is to bring the load position ( x L , y L ) above a defined inertial reference point ( x L,ref , y L,ref ) (see Fig. 5). For this task, the load motion relative to the helicopter must be minimized requiring a stabilization function. The automatic function means that the pilot does not control the helicopter directly but over secondary control inputs (e.g., beep commands). The need for an automatic function for load positioning was motivated by the inherent design conflict between HQs and load control when the pilot controls the helicopter manually and load motion is fed back to the rotor controls. This conflict becomes most apparent during a precision maneuver when the pilot tightly closes the control loop as during load handling in hover [12].

Load stabilization controller
The ALDS was designed for two different helicopter response types Attitude Command (AC) and TRC. From the literature [16] it is known that by feeding back the cable angle ( c , c ) and cable rate ( ̇c ,̇c ), to the rotor controls the damping of the load pendulum motion can be increased effectively. This control strategy uses the helicopter motion to damp the load pendulum motion. In Fig. 6 the controller of the load stabilization is shown.
The sum of the cable rate and cable angle augmented by static gains ( K̇c ,1 ,K c ,1 ) form the signal that is fed back to the input signal of the active helicopter mode (i.e., pitch AC). Due to the aerodynamic drag of the slung load there will be a constant trailing angle in forward flight. The cable angle signal is processed with a washout filter to eliminate this non-zero steady state during forward flight conditions. In TRC mode different feedback gains ( K̇c ,2 ,K c ,2 ) are used and the resulting feedback signal is fed back to the velocity command input.

Load positioning controller
The load position controller is an extension to the TRC mode and calculates velocity commands to maneuver the helicopter to position the load over the target point. For this task, feeding back the cable angles and rates, which measure the load motion with respect to the helicopter, is not sufficient. Without feedback of a position signal either of the helicopter or the load, load motion feedback causes the helicopter to maneuver over the load to damp the load motion. Therefore, additional feedback of the load position ( x L , y L ) with respect to a geodetic reference position ( x L,ref , y L,ref ) is required to minimize the position error (e). The controller of the longitudinal axis is shown in Fig. 7. To minimize the position error, the difference between reference and measured load position is fed back with a proportional gain. Feedback of the load motion (i.e., cable angle and cable rate) is needed for stabilizing the helicopter and load modes. The load position ( x L ) and cable angle ( c ) are positive when the load is ahead of the helicopter (see Fig. 5) so that all feedback

Control system architecture
The ALCS uses a multi-loop structure shown in Fig. 8. The inner-most loop provides AC in pitch and roll, Rate Command (RC) in yaw and collective (not shown). It is used to stabilize the basic helicopter during manual pilot control. TRC is provided by a loop around the inner-loop with feedback of the inertial velocity in longitudinal and lateral directions calculating attitude commands in pitch and roll, respectively. Load stabilization is provided by the load controller ALDS under piloted control. In AC mode load motion feedback is provided by the load stabilization controller ALDS 1 . In TRC mode the feedback path of ALDS 1 is inactive and the feedback path of ALDS 2 is closed. The load motion sensor provides the inertial cable angles ( c , c ) and their rates ( ̇c ,̇c). Is the load positioning controller ALPS active, the feedback paths of the load stabilization are as depicted in Fig. 8. When the pilot controls the helicopter in AC or TRC mode by the input du and ALPS is engaged, the pilot is kept out of the control loop and the ALPS controller calculates velocity commands to minimize the error between reference and actual load position. In ALPS mode the input du sets the reference load position (i.e., the target load position). The actual load position in longitudinal and lateral direction ( x L , y L ) is calculated from the cable length and inertial cable angles ( c , c ).
The optimization of the overall control system was performed using CONDUIT [17]. The inner-loop and the velocity loop were designed as augmentation for the bare airframe (the helicopter without slung load).
The inner-loop was designed to meet requirements for the control system to achieve predicted Level 1 HQs, such as piloted bandwidth [18], stability margins [19], and disturbance rejection [20]. The velocity loop providing TRC was optimized with the gains of the inner-loop fixed. Further details regarding the design and optimization of the AC and TRC modes are described in Ref. [13].

Optimization of the load controller parameters
The control parameters of the ALDS and ALPS controller have been also determined using the multi-objective optimization of CONDUIT. The optimization has been performed for the ALDS controller and ALPS controller separately. From the optimization, three parameter sets were obtained: two parameter sets for ALDS (one set for each helicopter control mode AC and TRC) and one parameter set for ALPS. During the optimization of all three configurations the innerloop parameters were not defined as design parameters  so that they remained fixed. The optimization process in CONDUIT is driven by the design specifications [21]. The specifications used for ALDS and ALPS design are listed in Table 1.
The first three specifications are general control design criteria and used for both load controller. They ensure absolute stability of the system and minimum stability margins which are evaluated for the open-loop response at the actuator input of each control axis. The damping specification addresses the minimum damping of the system poles. The low value for the boundary between Level 1 and 2 was defined due to the low damping of the load pendulum modes which are characteristic for a helicopter with slung load, especially in low speed flight conditions [22]. The damping of the load modes can be achieved with load motion feedback but at the cost of reduced damping of the modes associated with the helicopter rigid body motion [16]. Therefore, the boundary for minimum damping was defined at a low value that allows the optimization to find a possible solution.

Optimization of the slung load stabilization controller (ALDS)
During hover when the ALPS is active, the pilot is kept out of the control loop. The pilot only gives commands indirectly via a four-way-switch on the cyclic stick. As the load stabilization provides load motion damping in low speed and forward flight under piloted control, the conflict between piloted handling and load control had to be accounted for. This was realized by adopting the idea of a task-tailored load control system presented in [4]. From this idea, the concept of switching between different load control modes depending on piloted control has been applied to the load stabilization system of this study. When the pilot controls the helicopter directly by moving the control stick, the load control mode that provides acceptable HQs is active. In this control mode the load damping is less effective as in the control mode that provides highest load damping when the pilot is not controlling the helicopter and no degradation in pilot handlings are apparent.
The two load control modes use different control parameters which are optimized by including the two specifications for ALDS design (see Table 1) in the specification set. The two specifications represent the conflict between HQs and load damping as they are competing design criteria. This means that improvements in one design criteria (e.g., increased load damping) can only be achieved at the cost of the other design criteria (e.g., degraded HQs).
The input for the optimization visualized in Fig. 9 are frequency responses between stick input and helicopter attitude angles as well as frequency responses between stick input and load cable angles for different models (e.g., bare airframe and helicopter with attached slung load with ALDS off/on). The output of the optimization visualized in Fig. 9 are optimized gain sets for the ALDS controller (e.g., long. cable angle gain K c and long. cable rate gain K̇c ). The CONDUIT specification called Disturbance Rejection Peak (abbreviation used in CONDUIT: 'DstPkG1') which is normally used as design criteria for the disturbance response calculates the maximum magnitude peak of the frequency response between control input and load cable angle (Fig. 9, right). In this frequency response, the magnitude peak appears at the load pendulum frequency and a low peak value is associated with a high damping of the load motion due to a control input.
The specification 'FrqSLP1' addresses the impact of the load dynamics on the helicopter response and, therefore, on the HQs. This user-written specification has been adopted  from the slung load HQs specification that was presented in [23] and has been used for the design of slung load control systems as in [4] and in [10]. The impact can be characterized by the distortion of the frequency response of the attitude due to a control input (Fig. 9, left). The distortion is described by the depth of the notch in the magnitude response and the frequency of the -135 deg crossing or the lowest phase value near the load mode.
The specification 'FrqSLP1' used in this study differs from the original specification as follows: 1. Calculation of the magnitude notch only: The specification is one dimensional. The depth of the notch in the magnitude response is calculated only. The -135 degree frequency is not taken into account as the influence of the notch depth on the distortion of the attitude response has been found to be higher. 2. No fixed boundary between Level 1/2: As the notch depth varies with the method of frequency response calculation (e.g., from linear model or system identification), the boundary of the original specification would have been too restrictive for this study. 3. Constraint setting as 'Objective': The original specification is used as 'Soft Constraint' which means that the optimization is not driven by this specification. Setting the specification as 'Objective' as in this study, the optimization algorithm tries to achieve a solution with the best possible result for this specification.
The aim of the optimization was to find two designs for the load stabilization controller: one design for good piloted handling with sufficient load damping ('Low Damp' design) and one design for maximal load damping ('High Damp' design).
The optimization process can be described with Fig. 9 which shows the calculation of the competing design specifications and the results for the two control designs in the CONDUIT specific HQ plots. To cope with the competing characteristic of the two specifications, the specification 'FrqSLP1' (Fig. 9, left) was set as objective specification and the specification 'DstPkG1' (Fig. 9, right) as soft constraint with varying Level 1/2 boundary. With this setting, the optimization tries to push the result of the specification 'FrqSLP1' as far as possible into the Level 1 region. This means minimizing the magnitude notch depth while ensuring that the results of all other specifications (e.g., 'Dst-PkG1') are in Level 1. For the 'High Damp' design, the Level 1/2 boundary of the specification 'DstPkG1' was set to 3 dB to obtain a design with high load motion damping. As an improvement of 'FrqSLP1' is accompanied by a degradation of 'DstPkG1', the result of the optimization is a design with the best possible solution for 'FrqSLP1', while 'Dst-PkG1' is just in Level 1 (i.e., on the Level 1/2 boundary). The result for the 'High Damp' design shows an effectively damped load motion response but at the cost of a significant increase in magnitude notch depth compared to the baseline design without load control.
For an improvement in 'FrqSLP1', the restriction of the minimum load damping requirements had to be relaxed by increasing the maximum allowable value of 'DstPkG1' which is realized by shifting the Level 1/2 boundary towards greater values (i.e., 12 dB). This leads to a design with reduced load damping ('Low Damp'). The result of this design shows a less damped load motion (Fig. 9, right) but also a less deep magnitude notch (Fig. 9, left). It can be expected that the less distorted attitude response of the 'Low Damp' design leads to a better piloted handling with active load feedback. The values for the Level 1/2 boundary of 'DstPkG1' were found after iteration of this value from 0 to 20 dB and analyzing the responses in the attitude and the load motion both in time and frequency domain. The boundaries have been selected on engineer's judgement. 3 dB means high load damping and an acceptable response of the helicopter's attitude. 12 dB means a smaller response of the helicopter attitude and less, but still acceptable, load damping.

Optimization of the slung load positioning controller (ALPS)
The ALPS optimization method is explained in Ref [13]. for the case of a helicopter with rescue hoist. The method can be in general transferred to the case of the central cargo hook. For the optimization of the load positioning control parameters all inner-loops of the bare airframe control were closed and their control parameters held fixed. The feedback path of the stabilization controller was open (see Fig. 8). In addition to the three specifications which address general design requirements, two specifications were used for the ALPS design (see Table 1): one specification ('DmpTmG1') that addresses the damping of the response of the load position and one specification ('CrsnMnG1') that addresses the crossover frequency of the load position loop and, therefore, the performance of the ALPS controller. In contrast to the design of the load stabilization controller, the damping of the response was calculated in the time domain to ensure that the overshoot in the response is taken into account properly during the optimization. A high damping in the load position is associated with a low overshoot in the response which is desirable for precise load position during load repositioning maneuvers.
The crossover frequency determines the quickness of the load position loop and a high value results in a faster load positioning. Load position damping and crossover frequency are competing design criteria. Either the load is highly damped so it takes long to reach the final position without overshooting or the positioning is fast and reaches the final position within a small rise time but overshoots the final position. Therefore, it would not be useful trying to maximize both specifications simultaneously. As a consequence, the damping specification was defined as 'soft constraint' with a minimum damping ratio. The specification for minimum crossover frequency was defined as 'Objective', and therefore, the optimization tried to maximize the crossover frequency of the load positioning loop meeting the requirements of all other specifications.

Automatic load stabilization-features and simulation results
The switching of the load stabilization between the modes for effective load damping ('High Damp') and piloted handling ('Low Damp') is triggered by the detection if the pilot controls the helicopter actively ('Hands ON') or if the stick is in detent ('Hands OFF'). Detent is the stick position for a trimmed flight condition indicated to the pilot by small breakout forces in the force deflection curve of the cyclic stick. Pilot action is detected when the control input exceeds the detent position by 2% of the full stick deflection range. In addition, the stick deflection condition must be met for 1 s to avoid mode toggling when the stick crosses the detent position during control reversal. If the conditions for 'Hands ON' are detected, the load control parameters of the 'Low Damp' design are active. During the conditions for 'Hands OFF' the parameters of the 'High Damp' design are active. This ALDS control system with the switch between the two load control modes is named as ALDS mode 'AutoDamp'. The switching between the two load control modes for load stabilization is implemented in the control system as an output blending. For the study, the ALDS control mode with the highest possible load damping will be evaluated. Therefore, the ALDS mode 'HighDamp' does not provide the automatic load control mode switch and uses the parameters of the 'HighDamp' design only. This configuration is used for comparison reasons. In Fig. 10 the offline simulation results for the load stabilization control in AC mode are presented. The automated test input is a doublet input. The load pendulum motion is excited. Without ALDS, the helicopter attitude follows the commanded response but an undamped load motion is apparent. With the ALDS mode 'HighDamp', the load motion is damped effectively but at the cost of a highly distorted attitude response. The ALDS mode 'AutoDamp' also shows a distortion of the attitude response but less pronounced. The load damping is slightly reduced compared to the design with maximum damping but still effective. The signal 'SLD Mode' shows the switching of the load control modes. The delay between pilot activity and change of the load control modes occurs due to the conditions of the pilot activity detection and due to the blending process.

Automatic load positioning-features and simulation results
The functions of the load positioning system can be explained with the offline simulation results of Fig. 11. Starting in the hover condition the pilot performs a doublet input. The load motion is excited and after 20 s the automatic load positioning is engaged (see 'ALPS Control Command') and the calculation of the load position begins. The ALPS mode starts in the load position hold mode trying to hold the load over the position at the time of engagement. As the change in the signal 'ALPS Control Command' indicates, the load control is engaged with a fading process to prevent an excessive helicopter response due to the control commands of the load positioning controller. The commanded attitude is automatically faded between the commanded attitude value when ALPS is disabled and the commanded attitude value when ALPS is enabled. The fading time is a function of the load motion at the time of engagement and increases with the amplitude of the cable angle. The ALPS controller is fully effective when the fading process is finished. After 60 s, the load repositioning maneuver is initiated by setting a longitudinal load reference position with the beep command. The load position follows the commanded repositioning in forward direction and at 90 s in backward direction in the same manner. With ALPS active, the load pendulum motion is also damped well.

Piloted simulation study
A piloted simulation study was conducted in DLR's AVES [24] with the advanced ALCS used for the ACT/FHS simulation model with central mounted cargo hook and a load

Experiment set-up
Three experimental test pilots (see Table 2) evaluated six different control law configurations using the Load Placement Mission Task Element [4] (see Fig. 12). AC and TRC without ALCS are the benchmark configurations. Direct comparisons can be made between these configurations with the ALCS designed for high load damping ('SLD_High') and the ALCS configuration with the automatic blending between the two gain sets to improve HQs ('SLD_Auto'). Each experimental test pilot evaluated each configuration. After several training runs with each configuration three evaluation runs for each pilot and configuration have been made to ensure repeatable performance. This result in one representative HQR based on the evaluation runs per pilot and configuration. For each test point in total 9 datasets are available for the performance analysis in Sect. 4. Tests were conducted in order of Table 3, starting with the configuration at the top.
The pilot completes a diagonal maneuver with a defined Groundspeed (GS) (see Fig. 13). In the presented study a higher GS was selected (12-16 kn) than originally proposed (6-10 kn) by Ivler et al. [4]. Latest studies by Bellandi et al. [11] showed that a higher GS is necessary to force a reasonable load swing during the deceleration and stabilization phase of the helicopter. The approach was flown in a height of 50-60 ft AGL. After deceleration and stabilization of the helicopter the pilot tries to place the load on the target marked on ground (see Fig. 12). In contrast to the original maneuver the pilot was not instructed by any crew member. A "camera" view of the load (image from simulated visual environment) was shown on the display giving the pilot direct view of the load during the final phase of load positioning.
The MTE setup in AVES is shown in Fig. 12. Figure 13 shows the view from above in a sketch. All relevant task performance parameters for this MTE are given in Table 4.
AVES was operated during the test as a fixed-based simulator. HQs and pilot workload were evaluated using the   Cooper-Harper rating scale [25] and NASA Task Load Index (TLX) [26], respectively. In addition, objective task performance data (e.g., position error and load set-down time) were measured and analyzed.

Human machine interface
In Figs. 14 and 15 the human machine interface for the ALCS is shown. Figure 14 shows the pilot's single display used in the simulation study. The display itself is subdivided in four parts. In the upper right side the primary flight display is shown. Below a camera view with the top view of the load is displayed. This camera view can be provided by the slung load sensor (see Fig. 4 sensor for H135) which is also used to measure cable angles and cable rates used in the control laws.
In this camera view a marker is visualized on the ground as a red dot. This red dot indicates the position of the cable suspension point (cargo hook or hoist) (see sketch in Fig. 14) on the ground. The pilot can use this information as an additional reference. This means he has to bring the red dot directly over the intended load target point in the camera view to position the load. When ALPS is active the pilot can command a load position using a 4-way-switch on the cyclic stick (see Fig. 15). The display on the lower left side can be used to observe the automatic positioning process. In this display the commanded load target position and the current position of the cable suspension point is indicated. The upper left display shows the pendulum angle and can be used to observe the performance of the ALDS.

Results
In the following sections the results of the piloted simulation study in terms of performance data and HQs related data (e. g. HQRs, pilot workload as well as pilot stick activity) are presented, analyzed and discussed in detail. Attain a controlled hover within X s of initiation of deceleration 10 s 15 s Maintain altitude during translation and hover within ± X ft 4ft 6ft Controlled set-down of external load within X s of hover 50 s 120 s Load set-down position should be within a box ± X ft larger than the footprint of the external load on all sides 3ft 6ft The load should have no perceptible drift at touchdown √ -

Performance
The completion of the Load Placement Maneuver, as shown in Figs. 16 and 17 (flown by the same pilot), consists of several phases: Acceleration to a target groundspeed (12-16 kn), deceleration to hover, stabilization of the helicopter in hover and afterwards the positioning and set-down of the load. In both figures several markers are used to indicate each phase of the maneuver. In Fig. 16 the result of a run in AC mode without ALCS is shown. Caused by the deceleration of the helicopter the slung load starts to swing. The load swing is even slightly unstable in lateral axis. Due to the load swing, a precise load positioning was not possible and the pilot could only reach adequate performance. Figure 17 shows a run in AC mode with ALCS active ('AC + SLD_Auto') flown by the same pilot. During deceleration of the helicopter the load also starts to swing but now due to the active ALDS the load swing is well damped. Shortly after the helicopter is stabilized in hover, the pilot activated ALPS. In this mode the pilot is now indirectly controlling the helicopter. With the 4-way-switch on the cyclic (see Fig. 15) he is commanding the load target position. The objective of the ALPS controller is to minimize the error between commanded load target position and measured load position as best as possible. Figure 17 shows that with ALCS active the load can be precisely positioned on the load target position marked on ground in the MTE (desired performance reached). In addition, it is important to mention that the final load target position is reached without considerable load swing.
The performance analysis is based on the load set-down time and the load position error. The load set-down time is defined as the time between reaching a stable hover condition with the helicopter and the moment when the load touches the ground. The position error is defined as the   16 Load placement maneuver, AC mode without ALCS distance between the load target position marking on the ground in the MTE and the actual load position at the end of the maneuver. Figure 18 shows the load position error (mean values and maximum and minimum values) for each control law configuration over all test runs. Figure 19 shows the values for the load set-down time. From the comparison between the mean values of each control law configuration (see Table 3) it can be seen in Fig. 18 that a more precise load positioning was possible for configurations with the ALCS active. The load position error is getting smaller and also the scatter in the data is reduced when the ALCS was enabled. Both AC and TRC mode based configurations with ALCS active reached better positioning performance.  Figure 19 shows similar load set-down times for all configurations with a slight trend towards faster load-set down times when using TRC with ALCS functions. In 'AC + SLD_High' mode the load set-down time is higher than with the benchmark configuration AC and the data shows large scatter. This can be explained as follows: In the benchmark configuration AC (without ALCS) the load was swinging nearly undamped. The oscillation was even slightly unstable in lateral direction (see Fig. 16). The pilot realized this and tried to set down the load quickly before the situation becomes even more critical thus the pilot accepted a certain load position error. With ALCS active the load is well damped and pilots show the tendency to use the given time defined by the performance limits to improve the load position accuracy. The large scatter in the load set-down time for 'AC + SLD_High' can be explained by the distinctive control conflict between load damping and piloted control inputs before activating the ALPS in hover when the 'AC + SLD_High' mode was used. Best overall performance, this means smallest position error and fastest load set-down, was reached in 'TRC + SLD_Auto' mode. The configurations listed in Table 3 are representing an increasing level of automation from AC (lowest level) to 'TRC + SLD_Auto' (highest level). In Fig. 18 it can be seen that a higher level of automation results in improved load positioning performance.
In Fig. 20 normalization of the load pendulum motion is plotted. Normalized means the load deflection over the overall task is integrated and normalized by the overall maneuver time. The normalized load deflection is a quantity used to assess the effectiveness of the load damping during the overall maneuver. High values indicate a strongly swinging load during the maneuver. Small values mean the load is nearly perpendicular hanging under the helicopter during the entire maneuver. The values of each control law configuration and over all test runs are plotted in Fig. 20. The comparison of the mean values shows a clear improvement of the slung load damping and a reduction in scatter when using ALCS based on both control laws of the helicopter AC and TRC. For 'AC + SLD_High' mode, a more effective load damping, this means a lower normalized load deflection value, could be expected as this control mode was designed for high load damping. However, the conflict between piloted control and load stabilization during approach and deceleration phase of the MTE resulted in a more wobbly helicopter motion and thus in higher normalized load deflection.
The observable trend of Figs. 18 and 20 and is similar. In general a higher level of automation is resulting in a better load damping performance and in a better load placement performance.

Handling qualities ratings
The HQRs obtained from all three test pilots are shown in Fig. 21. A comparison of the HQR mean values of the configurations based on the AC mode shows an improvement of HQs with ALCS active ('AC + SLD_High' and 'AC + SLD_ Auto'). The 'AC + SLD_Auto' mode clearly improves the HQs in comparison to the 'AC + SLD_High' mode. Overall, the HQRs are located in the Level 2 region.
The mean values show that the configurations based on the TRC mode are rated better or at least the same than the configurations based on the AC mode. One reason for this is that the suspended slung load is more stable when the helicopter is operated in the TRC modes (even without ALCS) than in the AC modes (see Fig. 20).
The test pilots explained after each evaluation run they would divide the Load Placement Task mentally in two phases and rate each phase individually. The first phase is Fig. 20 Normalized load deflection of all tests Fig. 21 Handling qualities ratings the approach and deceleration to a stable hover. The second phase begins with the stable hover condition and ends when the load is placed on the ground. Without ALCS active, they would rate the first phase in the Level 1 region but the second phase in the Level 2 region. The reason for this is that the undamped load swing and the lacking capability to dampen the load actively by their control inputs result in low accuracy in load positioning. Finally, the pilots rated the overall task clearly in the Level 2 region. With ALCS active, the HQRs they mentally assigned for the two sub-phases would reverse. This means during the first phase, the approach and deceleration, they felt disturbed in their manual control due to the unpredictable helicopter response caused by the load motion feedback to damp the load swing. As a consequence, they would rate the first phase now in the Level 2 region. When ALCS is active they would rate the second phase, when the helicopter has to stay in hover and the load has to be positioned, in the Level 1 region, because the load is now well damped and especially ALPS enables an easy and predictable handling with low pilot workload. When the load stabilization mode 'SLD_High' providing most effective load damping was used, pilots felt clearly more affected by the load motion feedback than when they used the load stabilization mode with automatic load control switching'SLD_Auto'. This result is independent of the underlying helicopter response type AC or TRC.
The comparison in Fig. 21 between TRC mode and 'TRC + SLD_High' mode shows that two of three test pilots gave the 'TRC + SLD_High' mode a worse rating. These pilots recognized PIO (Pilot Induced Oscillation) tendencies in general when they fly in a mode based on TRC. Their comments give hints that the underlying TRC mode in its current design was not agile enough for them in this task and they had troubles to find an appropriate piloting strategy especially during the deceleration phase of the maneuver. Thus they were trying to compensate for the low agility coming with TRC mode through increased pilot control inputs (i.e., high gain and control reversal inputs). When ALCS was active, especially with the load stabilization mode for effective load damping ('TRC + SLD_High'), this strategy resulted in an aggravated conflict between load control and manual control. As a result, the pilots had to put more effort to compensate the helicopter response due to load motion feedback. With 'TRC + SLD_Auto' this conflict could be reduced and pilots rated the advanced ALCS better because of the less pronounced control conflict.

Pilot workload
During the study, the pilot workload was evaluated using the NASA-TLX workload rating scale [23]. The results are shown in Fig. 22. For the configurations based on the AC mode, a clear reduction in pilot workload is visible when ALCS is active. Especially the 'AC + SLD_Auto' mode results in a significant workload reduction, because in this mode, the conflict between automatic load damping and manual control is mitigated. The pilot's need to compensate deficiencies in the system is reduced in this mode. This is later shown in detail in Sect. 4.4 based on an analysis of the pilot's stick activity.
For the configurations based on the TRC mode, an increased workload can be observed in the 'TRC + SLD_ High' mode compared to the benchmark configuration. The two pilots who rated the system worse in the HQR (described before in Sect. 4.2) also felt a higher workload. This can be traced back to the underlying TRC mode which these pilots perceived as too sluggish and PIO prone. But all pilots commonly rated the workload with the configuration 'TRC + SLD_Auto' lower than with the configuration 'TRC + SLD_High'.

Pilot activity
The HQR and workload ratings clearly show an improvement for the modes which use the automatic parameter blending ('AC + SLD_Auto' and 'TRC + SLD_Auto'). For the modes without automatic parameter blending ('AC + SLD_High' and 'TRC + SLD_High'), the pilots commented unpredictable helicopter response during manual control. With the automatic blending, the influence of the ALCS on the helicopter response was reduced and the predictability of the helicopter response to pilots manual control inputs improved as well.
The plots and the associated characteristic control metrics show for the load stabilization without automatic blending ('AC + SLD_High') an increased control activity of the pilot, which is characterized mainly by more, larger and faster control inputs. One important metric of control activity is the attack number, which gives the identified peaks in the stick input rate (see Figs. 23 and 24). Another important parameter is the cutoff frequency, which can be interpreted as the bandwidth with which a pilot forms a closed loop through its manual control [27]. A higher value in the cutoff frequency, therefore, means that the pilot controls with a higher frequency. The 'AC + SLD_Auto' mode results in less pilot activity and was evaluated better by the pilots than the 'AC + SLD_High' mode. The results verify that the automatic blending function ('SLD_Auto') should be used in piloted handling in combination with a slung load damping system to ensure improved HQs.

Conclusions
An Automatic Load Control System (ALCS) for cargo operations was designed with the aim to reduce the pilot workload, increase the damping of the load motion and to improve the load positioning performance. The system was designed for two different helicopter flight control modes: Attitude Command (AC) and Translational Rate Command (TRC). A function has been developed that monitors pilot control inputs. Dependent on the amplitude and duration of stick deflection, the feedback signal for slung load damping is automatically blended between two different gain sets ('SLD_Auto' mode) to provide improved HQs during manual pilot control and to provide effective load damping when the pilot is passive. In comparison to this control   system, the use of a design providing high load damping, independent of the pilot's control inputs, was investigated ('SLD_High' mode). A piloted simulation study was conducted to evaluate different ALCS configurations with a simulation model of DLR's research helicopter ACT/FHS with a slung load. Three test pilots evaluated the system in six different control law configurations with increasing level of automation using a Load Placement Mission Task Element.
The main conclusions of the study are: • In benchmark evaluations for AC and TRC mode without the ALCS, the slung load starts to swing during the load placement maneuver. Pilots were only able to reach adequate positioning performance and experienced high workload during the task. • With the ALCS active, the load swing was well damped and the load could be placed precisely and fast using an automatic load positioning function. Pilots experienced low workload during hover and load set down. A trend could be shown in the study that a higher level in automation is resulting in a better load placement performance as well as improved load damping. • During the approach and deceleration phase of the load placement maneuver, pilots commented unpredictable helicopter response to manual control inputs when the automatic load stabilization in 'SLD_High' mode was active. Especially when using the system in the underlying TRC mode the pilots commented that the system feels sluggish and thereby is prone to pilot induced oscillations. • The 'SLD_Auto' mode that provides the automatic parameter blending between two gain sets was rated better by the pilots than the 'SLD_High' mode. Handling Qualities (HQs) could be improved and pilot workload was remarkably reduced in 'SLD_Auto' mode. Pilots commented that the helicopter response during manual control in the approach and deceleration phase of the Load Placement Task is in this mode much more predictable. Analysis of the pilot's stick activity confirmed a reduced pilot compensation.
From all results presented in the paper (i.e., performance data, Handling Qualities Ratings and workload data) it can be concluded that there is a clear tendency that a higher level of automation is resulting in a better performance in load handling at low speed and hover. This means a faster and more precise load positioning and increased load damping but simultaneously also the HQs can be improved and pilot workload can be reduced remarkably by the use of ALCS, especially if the load stabilization modes with automatic mode switching are used. This shows that the ALCS technology is able to contribute to high level objectives, such as increasing flight safety and effectiveness of slung load operations.