SARA Questions & Answers

Please submit your questions via email to eric.spero.civ@army.mil and brett.h.piekarski.civ@army.mil.

Comments or questions submitted should be concise and to the point. In addition, the relevant part and paragraph of the FOA to which a comment or question pertains must be referenced. Responses to non-proprietary questions received will be posted to the SARA Program website under the “General Information/Questions & Answers” section for the benefit of all interested parties. All clearly identified and marked proprietary questions submitted will be responded to via an individual email response, not posted to the SARA Program website.

Below is an archive of general questions and answers, listed in order received.

When reading the two sprint subtopics, it seems that they are related more to ground vehicles than to aerial vehicles. Are ideas associated with aerial vehicle still welcome? (February 14, 2021)

SARA sprint cycle #2 is focused on autonomous complex terrain maneuver for ground vehicles.

Who were the teams selected for Sprint #1? (February 18, 2021)

The awardees for Sprint #1 are listed on https://www.arl.army.mil/business/collaborative-alliances/current-cras/sara-cra/sara-overview/, under the heading “Awards to academia and industry”.

Is it expected that the proposing team has a platform similar to ARL Warthog including compute and sensor payload? Can you please elaborate on different options about it? (February 19, 2021)

Recipients are expected to perform development, integration, and experimentation on their own local testbed platform, which does not have to be a Warthog UGV. Experimentation events at Graces Quarters and other ARL venues, as necessary, will take place on an ARL Warthog UGV. It is not a requirement for Recipients to match the ARL hardware stack such as power, compute, sensing, and network hardware, on their own testbed platform but research plan should address a pathway for transition of research solutions onto the ARL testbed.

Is it possible to get the ARL platform as GFE on the task? (February 19, 2021)

ARL researchers will test algorithms on local, ARL-owned UGV platforms. ARL UGV platforms will not be loaned out or otherwise made available for Recipients to work on at their own places of performance. Recipients are expected to perform development, integration, and experimentation on their own local testbed platform.

Even if the proposed team already has a similar platform, could the proposed budget include acquiring a platform, sensor and computer suite specifically for this effort? (February 19, 2021)

Within reason, it is acceptable for a proposing team to include a base robotic platform, sensor, and/or computer suite, specifically for this effort, in their program cost estimate.

Can you provide any detail on potential number of awards? The 1.9M total across awards is clear. (February 19, 2021)

We do not know ahead of time the number of awards that will be made. The number of awards depends on the quality, number of proposals, and cost of proposals that come in.

Are PIs asked to develop methods to identify/classify the type of terrain? Or is the type of terrain assumed to be known? (February 19, 2021)

It depends on the type of capability being proposed to add to the autonomy stack. Experimentation will primarily occur under circumstances where there is limited or no information about the environment ahead of time. Proposers should clearly identify the assumptions required for their contributions to the system.

In terms of simulation, can it be assumed that proposers will have access to your simulation environment? In addition, how open for modifications would it be? (February 19, 2021)

As mentioned in the webinar, the simulation environment is an active, living capability that will be available and is modifiable by Recipients. Environment models, platform models, sensor models, and new feature implementation are all eligible areas of contribution during execution of the Sprint 2 base period. All modifications and new features will be done in coordination with ARL.

Are there citizenship requirements for all developers? (February 19, 2021)

No, not at this time.

What is the typical award amount? (February 19, 2021)

There is not a typical or target award amount. In Sprint 1, Awards ranged from roughly $100K to over $700K, dependent on the level of effort proposed, material, equipment, and type and quantity of personnel, among other factors.

For cycle #2 the focus is primarily on the improvement of the perception component…is there any interest on the navigation component? (February 19, 2021)

The focus of Sprint 2 is not solely on perception. We are interested in end-to-end performance in the described scenarios. Proposing teams may address any part of the ARL Autonomy Stack.

Is access to the full stack available ahead of the proposal? (February 19, 2021)

No. In order to access the ARL Autonomy Stack, or the Unity simulation environment, an active agreement (e.g., Award, Contract, Grant, etc.) must be in place.

Will the webinar slides be posted? (February 19, 2021)

Yes, the webinar slides will be posted to the ARL SARA website.

Is there a description of the stack’s current capabilities, gaps and improvement priorities? (February 19, 2021)

We do not have a description of the stack’s current capabilities, gaps, and improvement priorities, other than what is in available in the Funding Opportunity Announcement and presented in the webinar slides.

How will the information about bounded corridor be provided to the robot? (February 19, 2021)

A ROS topic with poses and radius will be sent in a transform library (TF) frame that is known by the robot.

Is there a way to get engine RPMs or motor current measurements from the ARL robotic testbed platform? (February 26, 2021)

The platform reports left and right drive current, voltage and velocity as well as main battery voltage and current status. RPM is not explicitly published but can be determined.

Are there suspension travel sensors on the ARL robotic testbed platform? If not, can they be installed prior to the May 2021, October 2021, and April 2022 operations? (February 26, 2021)

There are no suspension travel sensors on the platform currently. The base platform and hardware stack envisioned for use in May 2021, October 2021, and April 2022 operations is as-described in the Sprint 2 Webinar slides and in the FOA. Throughout the Sprint 2 base period of performance, ARL expects to work closely with Recipients to address any approach-specific modifications as necessary. If modification is required to the base platform or the hardware stack in support of a proposed research effort, then the SARA collaborator will provide an overview of the modifications in their proposal. It is expected that Recipients will fund all hardware and software development, fabrication, and integration on their own local testbed, and will collaborate with ARL to ensure ARL receives hardware designs and software/drivers required to run additional or modified sensors on the ARL robotic testbed platform.

Is it possible to modify the ARL robotic testbed platform, for example by attaching a force sensing bumper to the front of the platform? (February 26, 2021)

The base platform and hardware stack envisioned for use in May 2021, October 2021, and April 2022 operations is as-described in the Sprint 2 Webinar slides and in the FOA. Throughout the Sprint 2 base period of performance, ARL expects to work closely with Recipients to address any approach-specific modifications as necessary. If modification is required to the base platform or the hardware stack in support of a proposed research effort, then the SARA collaborator will provide an overview of the modifications in their proposal. It is expected that Recipients will fund all hardware and software development, fabrication, and integration on their own local testbed, and will collaborate with ARL to ensure ARL receives hardware designs and software/drivers required to run additional or modified sensors on the ARL robotic testbed platform.

In both the SARA Sprint 2 webinar and the FOA, it shows the ARL robotic test platform to have an Ouster OS1-64. Ouster has indicated that, sometime in 2021, they would be providing a dual- or multi-return capability from this sensor via a firmware upgrade. Will the ARL testbed Ouster OS1-64 have the dual/multi-return capability when the Ouster OS1-64 is used in the May 2021, October 2021, and April 2022 events? (February 26, 2021)

ARL is in the process of upgrading the Ouster ROS package in the ARL Autonomy Stack. ARL does not have additional insight into the Ouster development timeline.

Will Recipients be able to control the Exposure/Gain settings of the Blackfly S cameras through the ARL Autonomy Stack? (February 26, 2021)

Yes. Currently supported reconfigurable parameters are roughly the same as those shown here: https://github.com/ros-drivers/flir_camera_driver/blob/kinetic-devel/spinnaker_camera_driver/cfg/Spinnaker.cfg

Will the focal length on the camera lenses be secured to not move (during robot movement through complex terrain)? Further, will Recipients have the ability to intrinsically calibrate the cameras if we want to change the focal length? (February 26, 2021)

The focal length on the camera lenses will be secured during robot movement. Changing focal lengths of cameras impacts the sensor intrinsic, and system-wide extrinsic, calibration. Therefore, Recipients will have the option to request focal length changes but since this request impacts other components as well as cross-platform reproducibility, a decision will take this and other factors into consideration (e.g., impact to functionality such as refocusing, impact to operations such as time and ability to recalibrate, etc.)

Will Recipients be able to shield the cameras from bright sunlight like a baseball cap bill? (February 26, 2021)

ARL can work with Recipients to come up with a temporary form/fit/function solution to be implemented in the field, as long as it does not interfere with the LiDAR or any other sensors.

From the SARA Sprint 2 webinar and the FOA, the ARL robotic testbed platform shows multiple imaging sensors, including FLIR Blackfly S cameras. The Sony IMX287 internal CMOS imaging chip inside these cameras is sensitive in the near-IR. First, do you know if the Blackfly has internal filters to block out near-IR? Secondly, if there are not internal filters, will it be possible for Recipients to use a near-IR band filter on at least one of the cameras to be able to grab imagery in the near-IR? (February 26, 2021)

The FLIR Blackfly S cameras on the ARL robotic testbed have the internal IR cut filter intact. If Recipients desire to modify the setup, for example using external IR cut filters and/or IR bandpass filters, then a reason, cost estimate, and description of manufacturer/model need to be included in the proposal. Overall, if modification is required to the base platform or the hardware stack in support of a proposed research effort, then the SARA collaborator will provide an overview of the modifications in their proposal. It is expected that Recipients will fund all hardware and software development, fabrication, and integration on their own local testbed, and will collaborate with ARL to ensure ARL receives hardware designs and software/drivers required to run additional or modified sensors on the ARL robotic testbed platform.

In the FOA, there is a “Chapter 1” and optional “Chapter 2”. Why is Chapter 2 labeled as “optional”? Is it because Chapter 2 is where optional follow-on work (up to 3 years additional) is described, while Chapter 1 is where the “sprint” initial 12-month effort is described? (March 2, 2021)

Chapter 1 of the proposal is where one describes their initial, or base period (also known as “seedling”) effort. Chapter 2 of the proposal is where one describes optional future follow-on work, which would begin after the conclusion of the base period, at the discretion of the Government. If one chooses to include a description of optional future follow-on work, then the time horizon for this work is up to 3 years beyond the conclusion of the base period. Please refer to the table titled “Anticipated Core Funding Cycles” on page 13 of 31 of the FOA for a depiction of the relationship between base period and option period.

For first-time proposers, the proposed effort should be planned to be finished in one year? In the proposed effort section, it is indicated that the section should include “a short description and justification for annual research goals of the proposed effort”, which implies that the effort expands for more than one year. Can you please elaborate on this aspect? (March 4, 2021)

The base period of Sprint 2 is expected to run for 12 months after receipt of award. Please refer to the “Proposal Intent” paragraph beginning on page 10 of 31 of the FOA, as well as the table titled “Anticipated Core Funding Cycles” on page 13 of 31 of the FOA.

Are Recipients allowed to use additional hardware during the evaluation? For example, the integration of a sensor not currently in the stack. (March 4, 2021)

The base platform and hardware stack envisioned for use in May 2021, October 2021, and April 2022 operations is as-described in the Sprint 2 Webinar slides and in the FOA. Throughout the Sprint 2 base period of performance, ARL expects to work closely with Recipients to address any approach-specific modifications as necessary. If modification is required to the base platform or the hardware stack in support of a proposed research effort, then the SARA collaborator will provide an overview of the modifications in their proposal. It is expected that Recipients will fund all hardware and software development, fabrication, and integration on their own local testbed, and will collaborate with ARL to ensure ARL receives hardware designs and software/drivers required to run additional or modified sensors on the ARL robotic testbed platform.

In the current stack, where/how is multi-modal fusion for terrain happening? For example, in the block diagram of the baseline autonomy stack, RGB and LiDAR data appear to take separate paths until the “Global Planner” block. Do they generate separate occupancy grids and/or cost maps? (March 4, 2021)

Currently, multi-modal fusion occurs at the costmap level; different pipelines produce custom costmap layers and those are used for navigation. This is in active development and performers should not consider themselves beholden to this framework. Our long-term goal involves tighter integration with the mapping pipeline via our message stream architecture.