Experimental Evaluation of Radar Sensor Mounting Height on Vehicle Active Safety Systems Performance

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Dear authors, the article is written in a clear manner and the presented results are clear. The English language does not need to be improved in my opinion. However, I would like to see some additional information in the article:

- What radar sensor was used?

- Is the radar coverage in Figure 2 given by the width of the main lobe for a decrease of 3 dB?

- The text could emphasize more that some of the curious values ​​can also be caused by the side lobes of the radar antennas? For example, CPT for a height of 100 cm with an MRR radar? But I am not sure if the side lobes are relevant in this article, since no information is given about the radar sensor.
- Could you please add to Figure 5 for an example for a few waveforms the theoretical dependencies resulting from equation (2)? It would help to validate that the test setup was close to ideal.
- You attribute the decrease in RCS for the corner reflector at short distances to insufficient illumination by the radar. However, keep in mind that a corner reflector with a side length of about 10 cm has a farfield of about 5 m and only from this distance does the RCS have an estimated 10 dBsm. When it is closer to the radar, the RCS is lower.

Author Response

 

 

 

1. Summary
Thank you very much for taking the time to review this manuscript. We appreciate the positive assessment of the clarity of the article and of the English language. We have revised the manuscript to provide additional information on the radar sensors, clarify the meaning of the radar coverage plot, discuss the possible influence of side lobes, improve the interpretation of Figure 5, and refine the explanation of the reduced corner-reflector RCS at short distances.
2. Point-by-point response to Comments and Suggestions for Authors
Comments 1: What radar sensor was used?
Response 1: Thank you for this comment. We agree that identifying the sensors only as LRR and MRR was insufficient. The manuscript has been revised to identify the tested sensors as a medium-range radar and a forward-facing radar. We also added a table with available specifications [line 137].
Comments 2: Is the radar coverage in Figure 2 given by the width of the main lobe for a decrease of 3 dB?
Response 2: The figure represents the nominal effective operational coverage used for interpreting the experiments, from the specification it is the width of the main lobe for a decrease of 3 dB. We clarified the caption and description of Figure 2.
Comments 3: The text could emphasize more that some of the curious values can also be caused by the side lobes of the radar antennas? For example, CPT for a height of 100 cm with an MRR radar? But I am not sure if the side lobes are relevant in this article, since no information is given about the radar sensor.
Response 3: We agree. We expanded the discussion to clarify that local non-monotonic or unexpected values may result not only from two-ray ground-reflection geometry, but also from the interaction between target scattering centers and the actual antenna beam pattern, including possible side-lobe reception. This is now discussed especially in relation to target-dependent behavior and the MRR child-pedestrian result around 100 cm mounting height.
Comments 4: Could you please add to Figure 5 for an example for a few waveforms the theoretical dependencies resulting from equation (2)? It would help to validate that the test setup was close to ideal.
Response 4: Thank you for this valuable suggestion. We agree that a theoretical reference based on Equation (2) would be useful for interpreting the measured RCS oscillations. However, after re-evaluating the available data and the assumptions required by the model, we concluded that adding theoretical curves directly to Figure 5 would not be sufficiently reliable. Equation (2) contains several sensor- and scenario-dependent quantities, which were not measured independently and are not fully available from the sensor documentation. As a result, any theoretical overlay would require strong assumptions and could be misleading if interpreted as a quantitative validation of the experimental setup. Therefore, we did not add theoretical curves to Figure 5. Instead, we revised the manuscript to clarify that Figure 5 should be interpreted as a qualitative experimental illustration of the multipath behavior predicted by Equation (2), rather than as a direct validation of a fully parameterized theoretical model.
Comments 5: You attribute the decrease in RCS for the corner reflector at short distances to insufficient illumination by the radar. However, keep in mind that a corner reflector with a side length of about 10 cm has a farfield of about 5 m and only from this distance does the RCS have an estimated 10 dBsm. When it is closer to the radar, the RCS is lower.
Response 5: Thank you for this important clarification. We revised the interpretation of the short-range RCS decrease for the corner reflector. The manuscript now states that the reduced RCS at short distances is caused not only by partial illumination and near-field field-of-view effects, but also by the fact that a trihedral corner reflector with an edge length of approximately 10 cm reaches its nominal RCS only in the far-field region. We added the Fraunhofer-distance explanation to support this interpretation [line 267].

 

 

Reviewer 2 Report

Comments and Suggestions for Authors

The paper addresses a relevant and practical question in automotive radar design, and the experimental setup is clear and well controlled. However, the manuscript is too descriptive in its current form and does not sufficiently explain why the reported trends occur.

The manuscript is clearly written in Sections 1 and 2, and I appreciate the detailed background and explanation of the measurement context.

However, the results section would benefit from deeper analysis and a stronger link between the observed trends and the underlying radar parameters, such as bandwidth, transmit power, and antenna beamwidth, since these directly influence detection performance and multipath sensitivity.

A lot of results are presented, but the discussion does not go far enough in linking the observed detection gaps and RCS oscillations to the underlying multipath geometry, target height, radar beam pattern, and height-dependent field-of-view effects. As a result, the reader is left with repeated figures and numerical trends but only limited physical interpretation.

A few specific points should be addressed:

- In Fig. 3d, the corner reflector appears almost indistinguishable.

- In Fig. 5, the label “CR | Heights” is confusing, because the plot appears to refer to radar mounting height, not the corner reflector.

- Figures 6 and 7 reveal that different target classes behave differently with mounting height. For example, for the LRR, the child pedestrian appears to perform better around 20–40 cm, whereas the adult pedestrian shows better continuity around 60–80 cm. Similar target-dependent behaviour is visible in the MRR results. However, no analysis or explanation of these results was given.

- Because of this, the combined interpretation in Figs. 8 and 9 is not fully convincing on its own, since averaging over different targets may hide important target-specific trends.

Overall, the paper presents valuable experimental data, but the discussion should be expanded to explain not only what was observed, but also how the radar parameters and target characteristics shape these outcomes.

Author Response

1. Summary
Thank you very much for taking the time to review this manuscript. We appreciate the positive comments on the relevance of the topic, the clarity of the experimental setup, and the controlled nature of the experiments. We agree that the original manuscript was too descriptive in the Results and Discussion section. Therefore, we expanded the discussion to better connect the observed detection gaps and RCS oscillations with radar mounting height, multipath geometry, radar coverage, possible antenna side-lobe effects, target height, and target-specific scattering characteristics.
2. Point-by-point response to Comments and Suggestions for Authors
Comments 1: The manuscript is too descriptive and does not sufficiently explain why the reported trends occur.
Response 1: Thank you for this important comment. We have expanded the Results and Discussion section to explain not only what trends were observed, but also why they may occur. The revised discussion now links detection gaps and RCS oscillations to the radar-target-ground geometry, vertical field-of-view limitations, antenna beam pattern, mounting height, target height, and target-specific scattering-center distribution.
Comments 2: In Fig. 3d, the corner reflector appears almost indistinguishable.
Response 2: Thank you for pointing this out. We improved the presentation of Figure 3d by changing the photo that clearly shows used corner reflector.
Comments 3:  In Fig. 5, the label “CR | Heights” is confusing, because the plot appears to refer to radar mounting height, not the corner reflector.
Response 3: We agree. The label in Figure 5 was revised to explicitly indicate that the plotted curves correspond to radar mounting heights.
Comments 4: Figures 6 and 7 reveal that different target classes behave differently with mounting height. For example, for the LRR, the child pedestrian appears to perform better around 20–40 cm, whereas the adult pedestrian shows better continuity around 60–80 cm. Similar target-dependent behaviour is visible in the MRR results. However, no analysis or explanation of these results was given.
Response 4: We agree that this required further discussion. We added a new explanatory paragraph discussing why the child pedestrian, adult pedestrian, vehicle target, and corner reflector behave differently. The explanation is based on target height, RCS level, scattering-center distribution, vertical field of view, and multipath geometry [line 281].
Comments 5: Because of this, the combined interpretation in Figs. 8 and 9 is not fully convincing on its own, since averaging over different targets may hide important target-specific trends.
Response 5: Thank you for this observation. We revised the interpretation of Figures 8 and 9. The manuscript now explicitly states that the averaged curves are intended only as system-level summaries and should not be used as the sole basis for selecting a mounting height. We emphasize that Figures 6 and 7 remain necessary for target-specific interpretation, especially for vulnerable road users.

 

 

Reviewer 3 Report

Comments and Suggestions for Authors

The paper evaluates how automotive radar mounting height affects detection continuity and RCS stability. The topic is important, but several methodological issues should be addressed before publication.

1. Sensor specifications: Section 3.1 identifies the sensors only as LRR and MRR. Please provide their manufacturer/model or anonymized specifications, including frequency, bandwidth, range and velocity resolution, angular resolution, field of view, update rate, and Tx/Rx MIMO configuration. The influence of antenna geometry and signal processing on the observed multipath effects should be discussed.
2. Data acquisition: Clarify whether the analysis used raw detections, point clouds, or tracked objects.
3. Alignment: Explain how roll, pitch, yaw, and boresight were verified or recalibrated after each height change.
4. Sensor height recommendation: Research question was probably addressed by sensor manufacturers. Are there any public data or general recommendations used by these companies? E.g. Continental, or Bosch.

Comments on the Quality of English Language

The English feels quite average and could still benefit from a thorough review.

Author Response

1. Summary
Thank you very much for taking the time to review this manuscript. We appreciate the detailed methodological comments. In response, we revised the manuscript to provide additional information on the tested radar sensors, clarify the type of radar data used in the analysis, describe the alignment verification procedure after each mounting-height change, and discuss the availability of public manufacturer recommendations related to radar mounting height.
2. Point-by-point response to Comments and Suggestions for Authors
Comments 1: Sensor specifications: Section 3.1 identifies the sensors only as LRR and MRR. Please provide their manufacturer/model or anonymized specifications, including frequency, bandwidth, range and velocity resolution, angular resolution, field of view, update rate, and Tx/Rx MIMO configuration. The influence of antenna geometry and signal processing on the observed multipath effects should be discussed.
Response 1: Thank you for this comment. We agree that the original description of the sensors as only LRR and MRR required clarification. In the revised manuscript, we removed manufacturer-specific identification and describe the tested units as anonymized medium-range and long-range automotive radar sensing units. To provide technical context without linking the study to a specific supplier or production part number, we added Table 1 with publicly available reference parameters from component-level and system-level automotive radar documentation. The table includes representative information for the 76--81 GHz FMCW automotive radar class, including available RF bandwidth, Tx/Rx channel configuration, transmit output power, receiver noise figure, and system-level field-of-view assumptions where publicly available. Parameters that are specific to the tested production units and are not disclosed, such as exact range and velocity resolution, update rate, complete antenna radiation pattern, side-lobe levels, and full Tx/Rx MIMO configuration, are now explicitly marked as not disclosed or configuration-dependent. The manuscript also clarifies that the experiments should be interpreted as an evaluation of complete radar sensing units rather than as a validation of a fully parameterized antenna or waveform model.
Comments 2: Data acquisition: Clarify whether the analysis used raw detections, point clouds, or tracked objects.
Response 2: Thank you for this comment. We clarified this in the Methods section. The analysis used raw per-scan radar detections. Tracked-object outputs from the radar tracking layer were not used. This was done to evaluate detection continuity directly at the detection level and to avoid masking short outages by tracker prediction or smoothing.
Comments 3: Alignment: Explain how roll, pitch, yaw, and boresight were verified or recalibrated after each height change.
Response 3: We agree that this information should be included. The manuscript has been revised to describe the alignment procedure after each mounting-height change. The radar orientation was rechecked and adjusted using the mounting rig to keep the sensor boresight aligned with the straight-line approach axis. Roll, pitch, and yaw were kept consistent across mounting heights, with an adjustment tolerance of approximately ±1° [line 188].
Comments 4: Sensor height recommendation: Research question was probably addressed by sensor manufacturers. Are there any public data or general recommendations used by these companies? E.g. Continental, or Bosch
Response 4: Thank you for this suggestion. We added a short discussion of this point. Publicly available supplier information generally describes radar functions, use cases, and broad performance capabilities, but does not provide a universal radar mounting-height recommendation applicable across vehicle platforms. Public target-validation procedures, such as Euro NCAP GVT radar-reflectivity measurements, define reference sensor positions and alignment conditions for specific measurement setups; however, these values are measurement-procedure conditions rather than universal vehicle-integration recommendations.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

I have read the revised manuscript and the authors' responses to my previous comments. The authors have addressed my concerns satisfactorily, and the manuscript has been improved accordingly. I am satisfied with the revisions and recommend that the paper be accepted for publication.

Author Response

We thank the Reviewer for the positive evaluation of the revised manuscript and for recommending its acceptance for publication.

Reviewer 3 Report

Comments and Suggestions for Authors

The revised manuscript has been substantially improved, and most of my previous comments have been addressed. However, I recommend a minor revision before acceptance.

The main remaining issue concerns the characterization of the tested radar sensors. Table 1 provides public reference parameters for the AWR1243 transceiver and ITU-R automotive radar classes, but these values do not characterize the actual MRR and LRR units used in the experiments. At the same time, the parameters most relevant to interpretation of the results—particularly vertical and horizontal field of view, update rate, bandwidth, angular resolution, and Tx/Rx configuration—remain undisclosed. Generic parameters of a reference transceiver cannot be considered equivalent anonymized specifications of the tested sensing units.

Please provide actual anonymized values or approximate ranges for the principal parameters of the tested configurations, particularly the elevation coverage used to construct Figure 2. The source or derivation of the boundaries shown in Figure 2 should also be stated. If disclosure is not possible, the unrelated reference hardware values should be removed and the study should be explicitly framed as a black-box case study of two production radar units.

The alignment description also requires a small clarification. A spirit level can verify roll and pitch but not yaw or boresight alignment with the approach axis. Please describe how yaw and boresight were checked and clarify whether the stated ±1° represents adjustment resolution, estimated alignment uncertainty, or measured tolerance.

Finally, please replace “raw detections” with “untracked sensor-level detections” and qualify the broad 40–100 cm recommendation in the Abstract as being specific to the evaluated sensors and controlled scenario.

Comments on the Quality of English Language

The English feels quite average and could still benefit from a thorough review.

Author Response

Summary: 

Thank you very much for the additional comments and for acknowledging the substantial improvement of the manuscript. We have revised the manuscript according to all remaining points raised in the minor revision. The main changes concern the characterization of the tested radar sensors, the source of the Figure 2 coverage boundaries, the alignment procedure, and the qualification of the mounting-height recommendation in the Abstract.

Comments 1: The main remaining issue concerns the characterization of the tested radar sensors. Table 1 provides public reference parameters for the AWR1243 transceiver and ITU-R automotive radar classes, but these values do not characterize the actual MRR and LRR units used in the experiments. At the same time, the parameters most relevant to interpretation of the results—particularly vertical and horizontal field of view, update rate, bandwidth, angular resolution, and Tx/Rx configuration—remain undisclosed. Generic parameters of a reference transceiver cannot be considered equivalent anonymized specifications of the tested sensing units. Please provide actual anonymized values or approximate ranges for the principal parameters of the tested configurations, particularly the elevation coverage used to construct Figure 2. The source or derivation of the boundaries shown in Figure 2 should also be stated. If disclosure is not possible, the unrelated reference hardware values should be removed and the study should be explicitly framed as a black-box case study of two production radar units.

Response 1: Thank you for this important clarification. We agree that the previous version of Table 1 could be misleading because it included reference parameters from generic radar hardware and standardization documents that did not directly characterize the tested MRR and LRR units. Therefore, we removed the unrelated AWR1243 and ITU-R reference hardware parameters from the sensor specification table.

The revised Table 1 now reports only the principal parameters of the tested radar configurations, using actual anonymized values or approximate ranges where available. Parameters that cannot be disclosed or are configuration-dependent are explicitly marked as not publicly disclosed. In particular, we added the approximate horizontal and elevation coverage values used to interpret the results and to construct Figure 2.

The manuscript has also been revised to explicitly frame the study as an experimental evaluation of two production radar sensing units, with the unavailable internal parameters treated as black-box properties of the tested configurations.

 

Comments 2: The alignment description also requires a small clarification. A spirit level can verify roll and pitch but not yaw or boresight alignment with the approach axis. Please describe how yaw and boresight were checked and clarify whether the stated ±1° represents adjustment resolution, estimated alignment uncertainty, or measured tolerance.

Response 2: Thank you for this clarification. We revised the alignment description to distinguish between the quantities that were directly verified and those checked geometrically. Roll and pitch were verified using an angle meter and a spirit level. Yaw and boresight alignment were checked geometrically with respect to the straight-line approach axis by positioning the target on the road centerline and driving the vehicle along this line directly toward the target. We also clarified that the stated (\pm 1^{\circ}) value represents an estimated alignment uncertainty after manual adjustment, not a directly measured angular error or adjustment resolution

 

Comments 3: Please replace “raw detections” with “untracked sensor-level detections”.

Response 3: We have revised the terminology throughout the manuscript. The phrase “raw detections” has been replaced with “untracked sensor-level detections” to avoid ambiguity and to better describe the data used in the analysis.

 

Comments 4: Please qualify the broad 40–100 cm recommendation in the Abstract as being specific to the evaluated sensors and controlled scenario.

Response 4: Thank you for pointing this out. We agree that the previous wording could be interpreted as a general design recommendation. The Abstract has been revised to clarify that the 40–100 cm range applies specifically to the evaluated radar configurations and the controlled straight-line test scenario used in this study.