Target deployment and retrieval using JIAOLONG manned submersible in the depth of 6600 m in Mariana Trench

China’s 7000 m manned submersible JIAOLONG carried out an exploration cruise at the Mariana Trench from June to July 2016. The submersible completed nine manned dives on the north and south area of the Mariana Trench from the depth of 5500 to 6700 m, to investigate the geological, biological and chemical characteristics in the hadal area. During the cruise, JIAOLONG deployed a gas-tight serial sampler to collect the water near the sea bottom regularly. Five days later, the sub located the sampler in another dive and retrieved it successfully from the same location, which is the first time that scientists and engineers finished the high accuracy in-situ deployment and retrieval using a manned submersible with Ultra-Short Base Line (USBL) positioning system at the depth more than 6600 m. In this task, we used not only the USBL system of the manned submersible but also a compound strategy, including five position marks, the sea floor terrain, the depth contour, and the heading of the sub. This paper introduces the compound strategy of the target deployment and retrieval with the practical diving experience of JIAOLONG, and provides a promising technique for other underwater vehicles such as manned submersible or Remote Operated Vehicle (ROV) under similar conditions.


Introduction
From June to July 2016, China's 7000 m manned submersible JIAOLONG (Liu et al., 2010;Cui, 2013) carried out Leg II of the 37th China Dayang Cruises (CDCs) on the southwest area of the Mariana Trench, about 200 km southwest to Guam (142°25′E,11°22′N). The diving area of this leg is on the south slope of the Challenger Deep, where the depth is from about 6300 m to 8300 m. This leg is funded by the Strategic Precursor Program of Chinese Academy of Sciences "Researches on Some Frontier Science and Technology Problems of Hadal Abyss", which targets the basic sciences of hadal biology, environmentology and geology. This includes the biological structure, the high pressure adaptation mechanisms, and the evolution of hadal biological community. It also focuses on the spatial and temporal characteristics of hadal circulation, the environmental effect of gas release from the sea floor, and also the plate subducting and serpentinization at subduction zone.
During the 2-month expedition, the JIAOLONG submersible has been deployed for 9 dives, completing numerous tasks that included: discoveries of the mud volcano on the south slope of the Mariana Trench, investigations on the geology, biology and geochemistry of this region; sampling of rocks (e.g. peridotite, gabbro etc.), sediment, macro-organism and micro-organism; and, in situ trap and fixation tests in the hadal area. The diving scientists include DING Kang, PENG Xiao-tong, WU Shi-guo, ZHANG Hai-bin, WANG Yong, et al.
JIAOLONG also deployed serial samplers to collect gas-tight seawater samples (we call it "Gas-Tight" for short) near the bottom on the north slope of the Mariana Trench. The Gas-Tight, developed by Wu et al. (2015), can collect water at regular time automatically. The water could be used to study the chemical variation at the bottom of the trench. JIAOLONG's tracks were obtained using the POSIDONIA Ultra Short Base Line (USBL) system developed by the IX-SEA Corporation (Zhu et al., 2014;iXblue Inc., 2015). The transponder for the system was on the back of the submersible and the array was installed on the bottom of XiangY-angHong 9, the mother ship of JIAOLONG.
It is usually believed that the USBL system is not suitable for an area deeper than 6000 m in actual applications. As the USBL array is installed on the mother ship, the large distance between the ship and the deeply submerged target could lead to the deterioration of the positioning accuracy. So the Long Base Line (LBL) positioning system is usually employed in depths of more than 6000 m. But the LBL system needs to deploy four sonar beacons separately to calibrate them and retrieve them one by one, which requires much more time and effort at sea. However, the USBL could be an option in deep areas. In the present study, the accuracy of the USBL system of JIAOLONG by using the practical data in the depth of 6600 m was analyzed, and a compound underwater positioning and searching strategy was proposed with five factors: the USBL data, the indicating marks, the sea floor terrain of the area, the depth contour, and the heading of the sub. With the help of that, other deep sea underwater vehicles such as manned submersibles or Remote Operate Vehicles (ROVs) could achieve the high accuracy in situ target deployment and retrieval with the USBL system only under similar conditions.

Target deployment
We deployed the Gas-Tight and five marks in the 118th diving of JIAOLONG and acquired the USBL positioning data via this process. We also analyzed the USBL data of the Gas-Tight and the marks and made some calibration using the sea floor terrain, the depth contour and the heading of the sub. The details are given in this section.

USBL data of Gas-Tight
During the 118th diving, JIAOLONG deployed the Gas-Tight and Mark 5, as shown in Fig. 1a, and collected some rocks, living beings and water in the same place, costing 85 min in all. The positioning data are plotted in Fig. 1b. The depth here is 6671 m, and the large distance between the sub and the ship led to the "gridding" phenomenon in the figure, which means the limit of the positioning resolution. The minimum grid along the longitude direction is about 10 m, which means that the USBL system of JIAOLONG cannot identify the difference within 10 m in such a deep location. The mean value and the standard deviation of this positioning process are calculated and shown in Table 1.

Marks deploying and the positioning result
Marks were of great importance when we used the US-BL in such a deep area, which we used to compensate the USBL system error. JIAOLONG carried out an exploration cruise on the Indian Ocean Ridge from late 2014 to early 2015, in which the mean distance deviation of its USBL system is 35.5 m; in addition, the underwater visual scope in the sub is about 8 m. So we decided to deploy marks at a distance of approximately 30 m. That means the submersible may get somewhere apart from the Gas-Tight about 35  m when we use the USBL system to search for it. From there, the diving crew could see a mark several meters away from them. According to the mark, they could find the Gas-Tight. The deployment of marks greatly increased the possibility to find the Gas-Tight. The plan of the mark deployment is shown in Fig. 2. Mark 5 would be together with the Gas-Tight, and the other four marks would be on the west, east, north and south of Mark 5, separately. Also, they could be on the northwest, northeast, southwest and southeast of it or moved partly according to the terrain and on-site situation. In fact, the locations of all the marks varied more or less in the actual deploying operation in Dive 118.
In the deploying operation, the submersible moved about 30 m (estimated by the diving team in the sub) from Mark 5 and Gas-Tight (M5&GT) to another mark according to a fixed heading value (more details in Section 2.5) in the plan, and deployed the mark with its manipulator, obtaining the USBL data simultaneously. After that, the sub went back to Mark 5 to confirm the distance and the direction. The process of the mark deploying using the submersible sample basket and manipulator is shown in Fig. 3. (2) Mark 1 was deployed on the northeast of M5&GT actually, and the distance to M5&GT obtained by the USBL data is larger than that of the actual situation; (3) Mark 2 was deployed on the south of M5&GT and the distance by the USBL data was shorter than that from the diving crew; (4) Mark 3 was on the northwest and its USBL data corresponded to the actual situation; (5) Mark 4 was on the east of M5&GT, and the USBL data were inaccurate. Table 1 shows the positioning data for the five marks, including the deployment duration, position coordinates and distance deviation. Deployment duration determines the amount of the positioning dots acquired on the same mark; and the distance deviation is the statistical standard deviation of the positioning dots involved and is strongly correlated with the duration. The more dots we got on one mark, the more precisely its mean value of the position coordinates becomes. As the submersible expended the most time on the M5&GT, we believe it is much more accurate than other marks. Therefore, we take M5&GT as a reference when we correct other marks' position in the following tests. Meanwhile, the distance deviation of M5&GT reached 41 m after more than one hour of measuring, so if the deviation of one mark did not exceed this value, the positioning result could be credible. Moreover, 41 meters could also be considered as an acceptable range of the following calibration.

Sea floor terrain
Sea floor terrain could be of great influence to the target   deploying operation. Gentle slope is the best option, in which the submersible can operate conveniently and safely while using the depth variation to assist its positioning. The gradient of the Gas-Tight deploying area in Dive 118 (shown in Fig. 5a) is about 10°, which is perfect in actual application. Compared with that, we planned to finish another deploying operation in Dive 122; but it became a big challenge to the submersible pilot. The north part of the area (shown in Fig. 5b) is greatly abrupt and sharp, the gradient of which is up to about 70° and acts as a danger to the submersible when it descends to the sea bottom; the south part, in contrast, was so plane that it gave the pilot little directional help. So it would be quite difficult to deploy and find something in this area. In fact, we deployed another Gas-Tight in Dive 122 and were going to retrieve it in the last dive of this cruise, however, the weather turned bad and we had to cancel the diving and return, leaving the Gas-Tight at the sea bottom.

Depth
The depth contour of the target area and the depth variation of the submersible are useful references to distinguish the direction at the sea bottom. When the investigated area is determined, the depth contour can be calculated according to the multi-beam measurement or other instruments. With the guidance of the depth variation, more information on positioning could be obtained. When deploying a mark, the depth of the submersible should be noted. Fig. 6 shows the depth variation of JIAOLONG in Dive 118 in the process of deploying five marks, which are of great value in the following positioning calibration based on the USBL data. When the sub searched the mark in Dive 120, the sub moved vertically to the depth contour to find the noted depth, and then along the depth contour to find the target.

Heading of the submersible
When the submersible moved from M5&GT to each mark, the diving crew should also note the heading of the sub. After the deployment finished, the sub turned around, changing the heading with 180° to confirm the direction and the distance along the coming path. Table 2 lists the directions of the four marks to M5&GT, which would be useful when the sub finds one mark in the searching diving.

Comprehensive strategy and result
To sum up, we proposed the comprehensive strategy to calibrate the position of the deployed marks and target: (1) Calculate the mean value of the coordinates of each mark acquired by the USBL. (2) Compare the heading from each mark to the target between the value from the USBL and that noted by the diving crews in the sub. If they do not match well, take the noted one from the crews as the standard and decide the calibration direction. (3) Compare the noted depth from the sub with the depth contour, combined with the estimated sub travelling distance from the diving crews, to decide the calibration distance and the calibrated position of the mark. (4) If the calibration distance is smaller than the error range of the USBL system, the calibrated Table 2 Heading of the sub recorded by diving crews from different marks to Gas-Tight  GAO Xiang et al. China Ocean Eng., 2017, Vol. 31, No. 5, P. 618-623 621 position is considered accurate; if the distance is larger than the USBL error range, we should reconsider whether the USBL-based calculation is correct, or whether the record from the diving crews is right, and then make a comprehensive decision. We revised Mark 1 to Mark 4 according to the method above as shown in Table 3. (1) Mark 1 is moved towards Mark 5, and the shift distance is about 40 m, equal to the USBL error range. (2) Mark 2 is moved along the opposite direction to Mark 5 based on the depth from the sub. (3) Mark 3 stays in situ. (4) Mark 4 is moved according to the record from the sub. Although it seems quite different in the relative direction with Mark 5, the absolute calibrating distance is within the USBL error range. Fig. 7 shows the final positioning result of five marks after comprehensive analysis.

Target searching and retrieving
JIAOLONG submersible retrieved the Gas-Tight in Dive 120 according to the calibrated result. The actual searching process is as follows: the destination of the sub was the position of M5&GT from the USBL when it fell from the sea surface. It descended from the east of the target and reached the sea bottom at the south of it. It then climbed up the slope along the vertical direction of the depth contour from the south to the west of the target. When it reached the same depth as the target, it moved eastwards along the depth contour and found Mark 3. At last, it searched the target with the navigation shown in Fig. 7 and found it soon, taking less than 20 min to finish the job successfully.
The submersible reached the same location in Dive 118 and Dive 120. The statistical analysis of the USBL data in the two dives is shown in Table 3, which shows that the straight-line distance deviation is positively correlated with the time length of the operation as well. The maximum value of distance is from M5&GT, which indicates that the absolute measurement deviation of the JIAOLONG USBL system could reach 47.5 m when it is at the same position 6600 m deep. The direct distance between the ship and the submersible is about 6700 m, and the ratio of the measured deviation and the direct distance is 47.5/6700-0.007. In other words, the accuracy of the USBL system is 0.7% of the direct distance. Also, 47.5 m approximately corresponds to   the distance deviation of 41 m, which shows the capability of the USBL system of JIAOLONG. The absolute deviation of 47.5 m is a challenge to the underwater operation in reality. However, other than the USBL data, we used the marks, the sea floor terrain of the area, the depth contour and the heading of the sub to calibrate and analyze in a comprehensive strategy, and finished the target search and retrieve task efficiently and successfully.

Conclusion
In the Leg II of the 37th China Dayang Cruises, JIAO-LONG manned submersible completed the target deployment and retrieval in the depth of 6670 m, which is the first time that scientists and engineers have finished this kind of task using the USBL system in such a deep area. We used a method involving the USBL system data, five position marks, the sea floor terrain, the depth contour and the heading of the submersible. The work we present in this paper provides a valuable method for other underwater vehicles in a similar scenario.