Scalable, Adaptive, and Resilient Autonomy (SARA) CRA
SARA Cycle 1 Questions & Answers
Below is an archive of general questions and answers, listed in order received.
Please confirm that notification will be given to an awardee by the DOD prior to any delivery of CUI, or if awardee deliverables will be considered CUI? (January 31, 2020)
ARL will notify an awardee of delivery of any Controlled Unclassified Information, awardee deliverables are not expected to be considered Controlled Unclassified Information.
Can we propose to add new connections between nodes that are not currently present in the ARL Ground Autonomy Software Stack? For example add a connection from “Position Estimates” in Perception Pipeline to “Behaviors” or “Local Planner”. (January 31, 2020)
Yes, new connections between nodes can be proposed as long as functionality of current code and nodes remains.
Will the training data for the object detection and terrain segmentation nodes be made available to train a new algorithm? (January 31, 2020)
Yes, we have the labeled data available for both of these tasks. Note that the quantity of data we have is more suitable for fine-tuning rather than complete training from scratch.
Can we get a list of what behaviors already exists in the system? (January 31, 2020)
Ground vehicle behaviors: Navigate to a point, Navigate to any point in a region, Navigate within a bounded corridor, Simple servoing to follow a tracked object/person, Navigate to a point with a cost-map learned from examples, Exploration to maximize geometric map information in target region, reduce map entropy of occupancy grid.
Do we maintain commercial IP for all the software generated as part of this program? Can the software being developed as part of this program be commercialized? (January 31, 2020)
Success of the SARA program is dependent on the establishment and maintenance of a collaborative research environment that encourages and facilitates the sharing of intellectual property while providing adequate protection of ownership rights. The United States Government (USG) shall obtain “Government Purpose Rights” to Intellectual Property developed in the course of performance under the Cooperative Agreement. “Government Purpose Rights” do not include commercial application and performers will be encourage to commercialize software developed. Proposals should address their intellectual property (IP) approach and how their approach will foster collaboration with ARL and other SARA Recipients, and how their solution will further advance the state-of-art of open source or government owned autonomy solutions. The SARA program has an emphasis on the transition of basic research to the military and other USG applications. Thus, the Participants will cooperate with ARL in transferring, or otherwise making available, rights to SARA Intellectual Property for USG purposes. Intellectual Property that is made or created by a Participant outside of the course of performing tasks under the Cooperative Agreement will be considered “Excluded Intellectual Property” and each Participant has the option to retain sole title and ownership to SARA Intellectual Property created solely by its employees. SARA Intellectual Property made or created jointly among Participants shall be jointly owned by such Participants.
How will Point A and Point B be specified? Will they be GPS locations? (January 31, 2020)
In an experiment with intermittently-available GPS, then we will use GPS coordinates. In a GPS-denied experiment, goals will be specified relative to the initial pose of the robot at the beginning of the experiment.
Will the government provide a GFE multirotor platform to integrate onto at the awardee’s facilities? (January 30, 2020)
At this point a multirotor platform is not planned to be provided as GFE to the awardee facilities and proposers should plan on having their own testbed for development. SARA developed GFE multirotor platforms will be available and used for integration and testing during the two-week experimental event in October. As the ARL testbed configuration will not be finalized until sub-topic #2 proposals are selected in order to ensure support of the proposed research, it is expected that the testbeds will not be built and ready for flight until some point into the program. At that point, depending on the number of awardees and the number of platforms available, some platforms may be provided as GFE to support preparation for the two-week experiment event.
Is the estimated funding of $3M just for the first year, or does that include optional efforts that would be selected from first year proposals? (January 28, 2020)
The estimated funding of $3M is just for the first year seedling efforts. Option funding will be from FY21 funds, subject to negotiation and is outside of the $3M allocation for Cycle #1.
Is the topic of audit trails (sometimes product logs) a relevant topic of research. These would seem to be helpful for After Action Reports and to analyze problems that emerge so as to accelerate continuous improvement processes. I’m suggesting something like a Flight Data Recorder with accompanying analysis software so as to retrospectively review performance. Has this been considered for autonomous systems programs, such as SARA? (January 28, 2020)
All the robotic platforms will use ROS and will have the capability to record bag files for trouble shooting, visualization of data, and creating data sets for off-line development. Therefore, audit trails or “Flight Data Recorder” like development is not requested in this sprint sub-topics #1 and #2. It may be relevant in sub-topic #3 as a component to a comprehensive overall system for managing, executing, and troubleshooting large-scale air-ground collaborative experiments.
Regarding Sub-Topic 3, we were not sure if the goal of this portion of the solicitation was to: create a test-bed, both in software simulation and in hardware, and then test algorithms on this test-bed; create or use existing robots that will work in a real forest environment; derive, prove and implement algorithms for multi-agent swarms that will enable multi-agent tasks in remote environments like a forest. Also, should the experiments, both in software and hardware, take place on a large number of robots. For the simulation, that makes sense, but for hardware, what is a reasonable number of mobile assets on which we should demonstrate. (January 27, 2020)
For Sub-Topic 3 we are looking algorithms and software to manage and control large robotic field experiments in both urban and forested environments. Per the FOA this also requires management of execution code, runtime monitoring, configuration management and version-control, and validation of emergency stop functionality and other safety measures. There is no simulation, hardware, or test-bed development as part of sub-topic #3. ARL is developing the simulation environment internally and outside of the SARA program. Any experimentation will only be to demonstrate functionality of the developed software. For an initial capability demonstration we will have two ground and two air assets working in collaboration as part of an ARL experiment during the April and October events that can be used for this purposes. All SARA sprint #1 experimentation will be on single platforms and could also be used. Future SARA sprints and experiments may have tens of both air and ground assets operating at same time over large areas.
Do you prefer one larger proposal addressing multiple component technologies or do you prefer separate smaller proposals for each component technology? (January 27, 2020)
No preference, it is dependent of technologies being proposed and proposer preference. If there are synergies between component technologies or they all fall within one Sub-Topic area then a single proposal is most likely best, if no synergies or connections between component technologies and they span multiple Sub-topic areas then separate proposals could be considered.
Do you prefer joint proposals with multiple institutions, each taking care of subtopic or a proposal from a single institution addressing one subtopic? (January 27, 2020)
No preference. Proposals can be from single or multiple institutions, but an award will be to a single entity so if multiple institutions are involved in a proposal, the subawardee relationship should be defined.
How do you expect operators/human users to work with the autonomy or other components? (January 27, 2020)
For this sprint we are only interested in increasing the autonomy of the platforms. The only human interaction would be through the ground stations, the number of interactions with the system to complete the missions will be monitored. Future sprints may consider more human-agent interaction tasks.
What kind of end goal mission would you test? (January 27, 2020)
Most testing will be simple point A to point B navigation tasks. There may be some area coverage tasks at point B.
Is it of interest to present solutions for obstacle collision recovery as a means to recover from navigation failure? (January 27, 2020)
That topic would be considered. While physical platform development is not a focus, any autonomy solutions that can increase resiliency of platforms in complex off-road outdoor environments where increased operational tempo may result in collisions would be considered.
How closely do you expect awardees to work with other awardees on this and other sprints? (January 27, 2020)
Awardees will be encouraged to work together but not required. Solutions will coordinated to run on ARL platforms during the two week experimental events and will require collaboration with ARL staff and some level of collaboration between all participants.
How closely do you expect awardees to be working with ARL personnel? (January 27, 2020)
SARA awardees will work very closely with ARL personnel. The two-week experiment events will be coordinated to align to planned ARL experimentation. SARA awardees will be running solutions on ARL platforms in collaboration with ARL staff and ARL developed solutions.
Will multi-agent extension be of interest as an option? (January 27, 2020)
Not for this sprint. Multi-agent topics may be a focus of future sprint topics. Options should focus on extensions for current sub-topics and proposed research.
What are your thoughts on radio solutions for inter-agent communications? (January 27, 2020)
For this sprint we will not be looking at inter-agent communications and radio solutions to support that can be included but not required.
Regarding topic 2: Can the existing implementation stack navigate without GPS? Also how detailed is the map given to the drone during the beginning of the mission for path planning? Is it a 3d map with co-ordinate of all the obstacles? or a 2d bird’s eye view? (January 27, 2020)
Yes, both air and ground platforms can navigate without GPS. Experiments will also look at intermittent GPS operation. In some cases no map will be given, some cases a 2-D bird’s eye view with stale obstacle information may be provided. A full 3-D map with detailed coordinates of obstacles will not be provided.
For the subtopic 2, is there interest to divert from the 2 sample platforms size and go at smaller scale in the order of 200-500g platforms? (January 27, 2020)
There is no specific desire to go to 500g or less platforms for this sprint. It is not expected that the SARA UAV test beds will be at that scale in order to support necessary sensor and processing needs but if desired capabilities can be achieved by proposed work it will be considered.
Will we be testing the system during the day, night, or both? (January 27, 2020)
We will be looking to test in all conditions so both day and night.
Would option funding come from the original $3M pool of funds? (January 27, 2020)
Option funding would come from FY21 funds, subject to negotiation and is outside of the $3M allocation for Cycle #1.
For sub topic 2, area of interest – “Architectures to support dynamic re-allocation of computing resources between subroutines, running within the Robot Operating System (ROS), based on changing task/mission context and on-demand information requests” should the solution be targeted for ROS or could it be for ROS2? (January 27, 2020)
ROS, future sprints may utilize ROS2 but all current solutions will be ROS based.
What language is the existing stack source code written on? (January 27, 2020)
Ground autonomy stack is on C++ and python.
Can the sensors be re-positioned on the GFE testbeds? (January 27, 2020)
Yes, proposers should indicate any additional sensors or sensor manipulation needed to support proposed research.
Is the intent of scaling for sub-topic 3 on both (a) quantity of platforms and (b) larger, higher value platforms? (January 27, 2020)
Yes. The intent is for experimental control software that can scale up in numbers of platforms. We are interested in controlling heterogeneous mixes of platforms to include both air and ground based systems. We do not expect to experiment with or control very large high value assets as part of SARA beyond the types of systems called out in the FOA.
For sub-topic 2, is there interest in developing autonomous behaviors to enable the UAS to launch, navigate relative to, and precisely land on a moving UGV? (January 27, 2020)
No, future sprints may address this but not for sprint #1. Sprint #1 is focused on single platform autonomy.
Roughly when will proposers be notified that an option has been accepted/funded? (January 27, 2020)
The actual date is still TBD but will be after the October experimental event.
For ground autonomy, are you interested in legged mobility platforms? (January 27, 2020)
No. We may consider legged mobility in future sprints but not for sprint #1.
Is it the intent of the government that contractors will be delivering source code on this program? (January 27, 2020)
Yes, the intent is that code will be added to the ARL software repository. An applicant may submit a proposal for IP protected solutions or solutions that do not provide source code, but applicants are reminded to review the evaluation factors as the applicant’s plan for source code will be evaluated as part of their collaboration plan.
Will this presentation be made available? (January 27, 2020)
Yes, will be posted to the SARA website.
What is the expected number of proposals that will be funded? (January 27, 2020)
There is no specific number of proposals the Government anticipates funding. The number of awards will be based on the quality of proposals received, the cost of those proposals commensurate with the level of effort, and the Government funding allotted for this program.
Can we submit a single proposal to multiple topic areas? (January 27, 2020)
Could you provide details on sub-topic #3? (January 27, 2020)
Per the FOA, we are looking for software solutions to run, control, and manage field experiments with heterogeneous mixes of air and ground platforms that can scale up to large numbers of platforms.
Is the language on page 15, Structure of the Award, to be interpreted that option funding is subject to negotiation and is outside of the $3M allocated for Cycle 1? Or, does proposed cost for the option period of 3 years to factor into the $3M total allocation? (January 27, 2020)
Option funding will be from FY21 funds, subject to negotiation and is outside of the $3M allocation for Cycle #1.
I missed the webinar on Friday Jan 24th. Is there a recording of it for me to view? (January 27, 2020)
We will be posting the slides from the webinar later today to this site, questions and answers raised during the webinar were recently posted.
How intermittent should be assume the GPS signals are for subtopic 1? Would the final GPS coordinates of the goal location be given? (January 24, 2020)
GPS will be scaled during experiments from full GPS availability to no GPS availability and in some cases be dependent of actual tree coverage and actual GPS signal connectivity. Final GPS coordinates will be given for goal locations.