Influence of Restrictions on Range of View From Cockpit in Operation of Hydraulic Excavator

We investigated the effect of restricting the ﬁeld of view from the cockpit on the operation of a hydraulic excavator. We developed a computer simulation for the turning operation of an excavator and we performed an experiment. In the experiment, we recorded the operation time required to complete various reaching tasks. We found that the operation time depends on the distance from the start position to the target position, target size, and the ﬁeld-of-view width. We propose an extension to Fitts’ law model to represent the operation time according to the operation’s index of difﬁculty, which consists of the target distance, target size, and ﬁeld-of-view width. The developed model was evaluated by the simulation results and the results obtained by an experiment with an actual hydraulic excavator. We observed signiﬁcant improvement in the coefﬁcient of determination between the operation time obtained by the experiments and the proposed index of difﬁculty.


I. INTRODUCTION
Hydraulic excavators are used in various tasks, such as excavation and dismantling, owing to their many degrees of mechanical freedom, and the variety of attachments that can be fit to the tip (Fig. 1). However, Akyeampong et al. [1] have reported that operator's burden is large because the human machine interface (HMI) of hydraulic excavators is difficult to use. The working process of a hydraulic excavator can be mainly divided into three steps: rotating, running, and excavating. To perform the rotating movements, the operator uses joystick levers located close to their hands to operate the hydraulic excavator. According to the international The associate editor coordinating the review of this manuscript and approving it for publication was Jenny Mahoney. organization for standardization (ISO) standard, the operators determine the rotational direction and speed of a hydraulic excavator by tilting the left lever toward the left and right directions.
In loading operations, the accurate positioning of a hydraulic excavator, such as moving the bucket just above a target, is important when performing excavation and rotating movements. This operation can be considered as a type of pointing operation.
In a conventional sense, a pointing operation refers to pointing or selecting a target on a graphical user interface (GUI) through a mouse or touch pad. Hayashi and Tamura [2] conducted pointing experiments with a bucket to verify the effectiveness of vibration for the joystick of a tele-operated hydraulic excavator system.
Pointing motions have attracted attention as fundamental kinetic characteristics of human beings, and various studies have investigated them over many years. For example, Woodworth [3] proposed a model wherein pointing motion can be divided into two motions, namely, planning time and adjustment time. Uno et al. [4] focused on arm joints during pointing motion and proposed a minimum-change model, wherein the total sum of rotational torque is minimized. Fitts [5] evaluated and modeled pointing motion with regard to the time required to complete a pointing task (pointing time).
According to Fitts' model, the distance from the starting position to the target, and the size of the target, affect the pointing time. This model is very simple and does not depend on the input environment or the input method. Many studies have employed this model to investigate interface enhancement. This model has also been used in the field of psychology. For example, Burno et al. [6] used the model to evaluate the performance of a three-dimensional (3D) gesture motion input interface, mouse, and touch screen. Kim et al. [7] proposed a system combining gaze and a non-invasive brain machine as the input interface and used the abovementioned model to evaluate it. In Fitts' model, only the distance from the start to the end of the target and the size of the target affect the time required to complete a pointing operation task with a hydraulic excavator. However, the operator's range of view also seems to affect the operation. Typically, the arm of the excavator located on the right side of the cockpit blocks the operator's field of view (Fig. 2). This means that, from the operator's perspective, the visibility from the right side is not as good as that from the left side. Kittusamy [8] proposed a checklist for the operation efficiency of hydraulic excavators. Kittusamy showed various items that affect operation efficiency, including items related to the range of visibility. But he did not mention how fieldof-view size affected work efficiency. The pointing time may be longer when the field-of-view restrictions are larger. Thus, the extended Fitts' model considers that the operator's range of view is more effective for expressing the rotational operation difficulty of a hydraulic excavator.
In this study, the effect of range-of-view restrictions on the rotational pointing operations of a hydraulic excavator was investigated, and a difficulty model for rotational pointing operations was developed. This paper is organized as follows: Section II introduces previously proposed pointing motion models. Section III describes the simulation conducted to measure the pointing time with different ranges of view, and presents the rotational pointing difficulty model that was developed based on the results. Section IV describes the experiments conducted with an actual human-operated hydraulic excavator to evaluate the validity of the proposed difficulty model. Section V presents the conclusions drawn from this study.

II. PREVIOUS STUDIES ON POINTING MOTION MODELS
Many studies have attempted to model the difficulty of pointing motion. One of the existing models is Fitts' law, which models the relationship between the pointing task's index of difficulty, ID, and pointing time, MT . According to Fitts' law, MT increases as the distance from the starting position to the target increases and the target size decreases (Fig. 3). Mackenzie [9] improved Fitts' model and expressed the ID as follows: where D is the pointing distance and W is the pointing size.
MT tends to linearly increase with respect to ID, which can be expressed using the experimentally obtained constants a and b, as follows: Fitts' law was originally proposed as a simple motor response model with one dimension. MacKenzie [10] extended the model to consider two-dimensional pointing operations. Subsequently, Murata and Hirose [11] extended it to consider three-dimensional pointing operations. Various studies have also extended Fitts' law to consider other effects in addition to distance and size. Jax et al. [12] proposed the following equation to predict the pointing time when a pointing motion is performed on a curved line, OI , and assuming that an obstacle exists: where a, b and c are experimentally obtained constants. Accot and Zhai [13] proposed an index of difficulty for the steering operation required by a vehicle passing an elongated path, such that the vehicle does not extend beyond the path's width (Fig. 4). MacKenzie and Buxton [14] proposed a formula that incorporates the distance L to the target, the target size W , and the VOLUME 8, 2020 target height H when extending Fitts's law to two dimensions.
Bi et al. [15] conducted a pointing test at the size of a smartphone and proposed a more accurate difficulty model for small displays. The operability of the pointing devices and environment can be evaluated by using appropriately designed pointing difficulty models.
To our knowledge, a pointing motion model that considers the range-of-view effect has not been proposed. The index of difficulty is expected to increase as the range of view narrows. However, it is not clear how this affects the index of difficulty. In this study, we developed a simulation environment for rotational pointing operations, and investigated the relationship between the range of view and the pointing difficulty. By using this model, we can evaluate and increase the rotate work efficiency of the excavator.

III. SIMULATION A. EXPERIMENT USING TURNING POINTING SIMULATION
To clarify the range-of-view effect and develop rotational pointing difficulty models for hydraulic excavators, we created system for simulating rotational pointing operations. Our simulation system consisted of a personal computer, monitor, and input joystick (Fig. 5). We created the simulation using Unity, which is a 3D game engine. The experimental subjects were observed on a screen and performed the operation using a joystick. The display showed the starting point, target, and target area (Fig. 5). The system and the tasks are shown in Fig. 6.  When the subject tilted the joystick, the camera turned according to the orientation and inclination of the joystick.
The task was completed when the operator stopped using the joysticks. The required rotational angle from the starting position to the target area, target size, and range-of-view width could be changed. We can change the field-of-view size of the camera in Unity. Fig.7 shows the display image when the field of view is changed. In the experiment, the time required to complete the task was measured and a five-second break was set between each task. When the subject tilted the joystick, the view was rotated at a rotational speed ω corresponding to the inclination of the joystick with a maximum rotational speed of 60 [deg./s] and a maximum mechanical angle of joystick θ max , as follows:  The participants performed the task three times under each condition. The order of conditions is random for each subject. We measured the operation time and calculated the mean time under each condition. The participants were 10 healthy males aged from 21 to 22 years old. We recruited the participants from the students at Hiroshima University. The measurement commenced after all participants had practiced the task several times. Informed consent was obtained from all participants prior to the experiments based on the Declaration of Helsinki. Table 1 lists all of the pointing conditions (V , D, W ), mean operation time of all participants MT , and standard deviation of time SD. Figs. 9 to 11 present the MT results when the target distance D, target size W , or range-of-view size V changed, while the other conditions remained the same. A one-way ANOVA was performed and found that the main effect of the size of the target and the size of the visual field are significant(F(3,12) = 86.143, p < 0.01, F(2,12) = 371.798, p < 0.01, F(3,12) = 289.732, p < 0.01). Based on Figs. 9 and 10, we confirmed that the time increased as the distance became larger or the size became smaller. These results can be explained by the original Fitts law. Johnson et al. [16] conducted experiments with a remotely controlled robot and showed that work efficiency depends on the field-of-view size of the presented image. These results are consistent with the results of our paper. Moreover, Fig. 11 shows that the time became longer as the range of view became smaller.

C. RESULT OF SIMULATION EXPERIMENT
The result revealed that the index of difficulty became higher when the target distance D increased, target size W decreased, or range-of-view size V decreased.

D. MODELING OF TURNING POINTING OPERATION
We added the following two conditions to the index of difficulty: when the target size W = 0 or the field-of-view size V = 0, the index of difficulty ID = ∞; when the target distance D = 0, the index of difficulty ID = 0.
We propose the following index of difficulty ID new that satisfies above conditions: Figs. 12 and 13 show the scatter diagrams and measurement time for the original index of difficulty ID (1) and proposed index of difficulty ID new (8), respectively. In these figures, the horizontal axis represents the index of difficulty, while the vertical axis represents the mean time for completing the task, MT . Table 2 lists the decision coefficients for each subject. We confirmed that, in the simulation environment, the performance of the proposed ID new was better than the performance of the original ID. VOLUME 8, 2020

IV. EXPERIMENT WITH ACTUAL HYDRAULIC EXCAVATOR
To test the applicability of the proposed index of difficulty to an actual environment, we conducted an experiment using a real hydraulic excavator with a limited range of view, in the same manner as in the simulation described in the previous section. Fig. 14 shows the hydraulic excavator (SK200-9, made by Kobelco Construction Machinery Co., Ltd.) that was used in the experiment. We placed two cones apart from each other and defined the interval between them as the target area. The participants were asked to operate the excavator from the start position by stopping the rod attached to the tip of the bucket within the target area using only rotational operation (Fig. 15). The rotational operation was carried out by tilting the operation lever in the cockpit. Other operations, such as running or excavating, were not used in the experiment. Partitions were set in the cockpit to change the operator's range of view.    Fig. 16 presents the method used to calculate the experimental conditions. The target distance D is defined as the required rotational angle from the starting position to the center of the distance between the cones. The target size W is defined as the angle calculated by the distance between the cones. The range of view size V is defined as the angle calculated from the range of view created by the partition. were considered. The participants performed the rotational pointing operation three times for every combination, and the operation time was recorded. The order of conditions is random for each subject. The participants were ten adult males aged from 21 to 52 years old. We recruited 10 participants from the employees of Kobelco Construction Machinery Co., Ltd., who has a pilot experience in the operation of hydraulic excavators. Table 3 presents the results obtained for each pointing condition, index of difficulty, mean operation time, and standard deviation.

C. EVALUATION OF PROPOSED MODEL
Figs. 20 and 21 show the experimentally obtained relationships between the operation time and the index of difficulty. Table 4 lists the coefficient of determination for all subjects.

D. CONSIDERATION
The mean determination coefficients for the proposed and original index of difficulty were 0.75 and 0.59, respectively. This indicates that the proposed index of difficulty, which considers the effect of the range-of-view from the cockpit, performed better in evaluating the difficulty of performing a VOLUME 8, 2020  rotational operation with an excavator, both in the simulation and in the experiment with an actual excavator. These findings can be applied to the design of a hydraulic excavator's interior. The cockpit of an excavator is typically shifted to one side, which results in different visibility for the left and right sides. According to the proposed model, the side with a limited range-of-view has higher operation difficulty, which makes it difficult to operate the excavator in an accurate and efficient manner.
Additionally, the proposed index of difficulty can be used to characterize hydraulic excavators. Card et al. [17] used parameters 'a' and 'b' in (2) to assess the performance of human-computer interfaces, and reported that the performance of the interface improved as the slope of 'a' became smaller. There are many types of commercial excavators with different sizes and power. Because the rotational speed of those excavators is also different, the parameters 'a' and 'b' in (2) are affected.

V. CONCLUSIONS
This paper proposes a rotational pointing difficulty model, which considers the range-of-view effect on the operation of a hydraulic excavator. To develop this model, we measured the operation time in a simulation environment with different target distances, sizes, and ranges of view. Based on the measurements, we propose a new index of difficulty. The proposed model was evaluated by conducting experiments with an actual hydraulic excavator. The experimental results revealed that the proposed index of difficulty has an average determination coefficient of 0.75 between the operation time and the proposed index difficulty.
In future work, we will carry out measurements with more participants and consider more parameters, such as the excavator operator's age and experience. Moreover, we will consider the effect of the interface's index of difficulty on the operating time. Finally, we will investigate the application of the proposed index to the design of a new excavator cockpits. Mr. Kurita is a member of the Japan Society of Mechanical Engineers, the Robotics Society of Japan, the Virtual Reality Society of Japan, and the Society of Instrument and Control Engineering. VOLUME 8, 2020