*5.2. Moving Platform Scenario*

In this scenario, three cases were analyzed when head-on, crossing, and overtaking motion. During the experiment, the ASV moved with a stable course and speed, while the inflatable boat moved according to the desired trajectory. Both boats were simulating typical collision situations that could reasonably occur while the ASV was deployed to collect measurements. In each of the three parts, the target was approaching its own ship from various relative bearings. In Figure 8, the situation is explained graphically, presenting the ranges of the relative bearing for each situation. It should be pointed out, that the terms head-on, corssing, and overtaking was slightly modified compared to the traditional understanding of IMO COLREG (Collision Regulations) requirements. For the purpose of this research, we proposed that head-on means the situation in which the target is approaching from the relative bearings in the range (−10◦;10◦); crossing means the situation in which the target is approaching from the relative bearings in the ranges (−90◦; −8◦) and (8◦; 90◦); and overtaking means the situation in which the target is approaching from the relative bearings in the ranges (80◦; −80◦). As it can be seen in Figure 8, the areas slightly overlap each other and the relative course of the object decides the type of movement. Such definition of the areas in the scenario ensures the analysis of the maximum detection range in the entire filed of view of the radar antenna.

**Figure 8.** Situation areas in dynamic scenario.

For head-on motion, the boats were moving toward each other from a large distance. The goal was to determine the maximum detection distance for such a boat. According to the declarative field of view, the head-on course should provide the maximum detection distance. The tracks analyzed in this scenario (after entering the field of view) are presented in Figure 9. The own ship was located in the beginning of the coordinate system and the x-axis is oriented with a relative bearing of 0◦. The relative tracks of the targets after entering the field of view are given.

**Figure 9.** Moving Platform Scenario tracks for head-on motion experiments.

The results for the head-on situation are compiled in Table 3. Because of the good convergence of measurements, it was decided that five iterations were good enough in this situation.


**Table 3.** Moving Platform Scenario results for head-on motion.

As shown in Table 3, the mean detection range for this type of boat is 169.04 m. This measurement was reproducible, such that according to the 3-sigma rule, the real detection range should vary from 168.5 to 169.5 m. This result generally confirms the declarative detection range for small cars onshore. The relative bearings confirmed the scenario assumptions (head-on situation) and the relative speed shows that the target approached at a nearly constant speed of 2.5 m/s.

In the crossing and overtaking motion, the main goal was to find the angles at which the target appeared in the field of view and then left it. Eleven tracks were recorded for both the crossing and overtaking motion experiments. The relative tracks are presented in Figure 8. The graphs show a plan view, in which the HydroDron is in the middle of the body frame coordinate system and the x-axis points toward the heading. The observed platform presented in Figure 6 (an inflatable boat) was moving according to the established patterns. The HydroDron was moving with a steady course and speed, while the target was maneuvering. As it can be seen in Figure 10a, the tracks were recorded in various distances, from a few meters up to 170 m (detection maximum for this type of target). In the case of overtaking movement (Figure 10b), only the moment of the first target detection is important

and thus only the incoming target was taken into account. Notably, the tracks were selected to verify the angles over various distances, both smaller and bigger. The measurement results for crossing the tracks are provided in Table 4.


**Table 4.** Moving Platform Scenario results for the crossing motion experiments.

**Figure 10.** Moving Platform Scenario tracks for (**a**) crossing and (**b**) overtaking motion experiments.

As shown in Table 4, the statistics for the crossing motion experiments are divided into the port side and the starboard side of the ASV (and radar). Additionally, the results are presented separately for the target coming into the field of view and then leaving it. The sample size was small due to the complexity of the study, so the T-distribution was used. In each case presented in Table 3, the mean value is within (40◦–45◦), which can be treated as the typical angular restriction of the field of view. However, the maximum values are more than 50◦ and the distribution is more or less symmetrical. Furthermore, the standard deviation and standard error achieved in these experiments are relatively big because the crossing motion occurred at various ranges. Based on evaluation of the detailed data, we found that at larger ranges, the angular field of view was smaller and the crossing target entered the view later; for example, when the target appeared at 169.2 m, the angle was −12◦.

The observations made in the crossing motion experiments were confirmed in the measurements for the overtaking situation, as shown in Table 5. Seven measurements are presented, together with statistics based on the T-distribution assumption. In these experiments, only incoming targets were analyzed, and the absolute value of the bearing was calculated without dividing it into portside and starboard side.


**Table 5.** Moving Platform Scenario results for the overtaking motion experiments.

Overtaking usually occurs at relatively small distances, a situation that was reproduced in this research. This results in a better and more accurate mean value of nearly 50◦. In the first measurement, where the distance was more than 60 m, the bearing was smaller. Summarizing these observations, the angular field of view should be determined as a function of range; for small ranges it is nearly linear (approximately 45–50◦), but for bigger ranges the field of view falls exponentially. To verify this hypothesis, the relative bearing graphs are presented in Figures 11 and 12. Figure 11 presents the relationship between the range and relative bearings, wherein the minimum- and maximum-recorded bearings for each distance are plotted. The envelope for more than 40,000 measurements collected in the Moving Platform Scenario is presented.

**Figure 11.** Relative bearing as a function of x-coordinate in the Moving Platform Scenario.

As shown in Figure 11, larger x-coordinates correspond to smaller bearings to detect the target. Although the maximum detection range is still approximately 170 m, the geometry of the sensor and experimental configuration suggests that targets at smaller distances will not be detected. For example, a target at a distance of 150 m at a relative bearing of 40◦ will not be detected by radar. This consideration leads directly to the detection area pattern presented in Figure 12. The graph shows the measurement points positions in the Cartesian coordinate body system. The vertical axis indicates the direction of

ASV movement with an envelope of detectable targets. The field of view appears as a quarter circle with a radius of almost 170 m. Up to approximately 120 m, the angular width of the field of view is almost identical (90–100◦); at further distances, the effective width is smaller. Notably, no lobes are observed, which could be expected, based on the declarative beam pattern. The envelope presented in the graph is generally a smooth line that was created based on the minimum and maximum range values for each bearing with a resolution of 1◦. A comparison with the measurement points indicates that this envelope is rather optimistic and the effective field of view is narrower. Some disturbances to the envelope can be observed at distances of about 100 m (x-axis), which are larger on the starboard side, where the graphed line is less smooth. One reason for this might be the mounting on the boat left of the echosounder pole. This hypothesis should be verified in the future.

**Figure 12.** Empirical detection pattern based on measurements collected in the Moving Platform; measured points are in green and the envelope is shown with a red line.

#### **6. Conclusions**

The research in this study provides an empirical analysis of surface target detection in a water environment with automotive radar, which can be used for the future development of tracking and anti-collision systems for ASVs. The research focused on identifying the detection ranges and field of view for various targets. Typical objects that could be met in the water environment were analyzed, including a boat and floating objects.

The overarching goal of the research was to verify a novel approach for object detection in a water environment. The novelty was based on using radar sensor for this approach, which is usually implemented in cars for road situations. This approach may in the future overcome the disadvantages of other systems used for anti-collision in ASV, namely laser rangefinders, lidars, and cameras. The proposed system was verified in real conditions with the online recordings.

The research showed that the system was capable of detecting many small targets but some objects, such as a fender, were not detected. Therefore, detection depends both on the size of the target and the material. In general, objects that are air inflated, such as fenders or airtoys, show worse detectability than solid targets, such as lifebuoys. Detection can be significantly improved using a radar reflector, however these reflectors are not usually deployed in practice. In general it can be said that the maximum detection range of small targets is about 15 m, while in very short distance (less than 3 m in research configuration) they are in the shadow due to antenna mounting. It can be assumed that 15 m is a reasonable distance to perform hard anti-collision maneuvers (like full stop) for such a small target with good maneuverability, however this judgment has to be verified in future research.

The second part of the study was conducted using an inflatable boat as the model object in motion. A complex analysis of the field of view for this target was performed, including the radial distances at different angles for various movement parameters. In general, the empirical research confirmed the product limitations and performance declared by the producers, under the assumption that an inflatable boat could be treated as a small car based on their size similarities. The maximum detecting range, confirmed with statistical post-processing, was about 170 m, while the field of view was about 100◦ (50◦ for each side). These values seem to be reasonable for planning anti-collision maneuvers with moving targets.

In summary, for larger targets that represent the greatest risk, the radar system provides good detection for anti-collision purposes. For smaller targets, the detection ranges are smaller, although for most targets, it would be small enough for the ASV to maneuver around. Additionally, some small targets were not detected. Generally, the automotive radar system may be a good basis for an ASV anti-collision system; however it should be supplemented with the integration of additional sensors, such as laser rangefinders. In the future, the detection stability and additional small targets should be investigated. It would be also interesting to see how this kind of radar would react in a sea environment. ASVs used at sea might be also a possible target of implementation.

**Author Contributions:** Conceptualization, A.S. and W.K.; methodology, W.K.; bibliography review, A.S.; acquisition, analysis, and interpretation of data, W.K., D.G-S. and W.M.; writing—original draft preparation, W.K.; writing—review and editing, A.S.

**Funding:** This study was funded by the European Regional Development Fund under the 2014-2020 Operational Programme Smart Growth; as part of the project, "Developing of autonomous/remote operated surface platform dedicated hydrographic measurements on restricted reservoirs" implemented as part of the National Centre for Research and Development competition, INNOSBZ and under grant No 1/S/IG/16 financed from a subsidy of the Ministry of Science and Higher Education for statutory activities.

**Conflicts of Interest:** The author(s) declare(s) that they have no conflict of interest regarding the publication of this paper.
