DARPA/MTO BAA HR001117S0043 Frequently Asked Questions Updated October 3, 2017

Transcription

1 DARPA/MTO BAA HR001117S0043 Frequently Asked Questions Updated October 3, 2017 Q1: Can a TA1 solution rely on intimate protocol or waveform knowledge, e.g. how to demodulate the signal? A1: Per the BAA, during evaluation, the government team may train and evaluate your system with additional waveform types not included in the previously distributed dataset. Q2: Should learned features include only the RF signal, or also include characteristics of the signal collection environment? A2: The learned features should emphasize the RF signal and be robust to environmental variations. There is no requirement to detect or label environmental characteristics, but this additional capability would be positive. Q3: Can you provide examples of important? A3: The definition of importance can vary across problem sets. An incomplete list of examples are provided below. During development of your system you may create your own definitions of importance for testing. 1. Signals that are anomalous relative to the majority of other signals present 2. The presence of a repetitious GPS signal amongst non-repetitious GPS signals 3. The presence of a mobile WiFi access-point 4. The presence of communications signals in a band where only radar signals are expected (or vice versa) Q4: Does the TA2 system need pre-programmed knowledge of common signals to understand salience (e.g. radar signal in a communications band)? A4: No. Q5: Can we propose to re-train the Task 2A system under several Task 2B control settings? A5: Yes. Q6: Will sample GFI data be available for proposal writing? A6: No.

2 Q7: Is the data classified or are there any other restrictions on the data? A7: The data is unclassified. It is unknown at this time whether any other restrictions, such as For Official Use Only (FOUO) will apply to the data sets. It is our desire to create a public dataset. Q8: How will datasets be disseminated? A8: At this time two dataset dissemination options are being considered: 1) Download via the internet 2) Distributed via portable (i.e. shippable) hard drives Q9: Will TA1 datasets include data gathered over non-consecutive days, under different conditions, using different equipment, and across different locations? A9: Yes. Q10: Will the training data labels have errors? A10: Yes, but no intentional errors. Q11: What type of equipment will be used to record GFI datasets? A11: The type of equipment used cannot be specified at this time. Q12: What labels beyond important will be provided on TA2 GFI data? A12: It is unknown at this time, what additional information will be provided in the in the TA2 GFI dataset. Q13: For TA2, will signals from multiple antennas be available simultaneously? A13: Yes, the TA3 provided hardware is expected to have multiple phase coherent receivers. Q14: Will the training data set include all of the evaluation data? A14: The government team will withhold some data specifically for the purpose of evaluation. Q15: What does streaming data mean during online evaluation?

3 A15: Streaming refers to samples which are provided in real-time from a device such as a RF receiver. It does not mean that the data is hosted online. Q16: Can the real-time constraint on Task 1A be interpreted as latency? A16: Yes. Q17: Without a SWAP limit on computation, why is achieving real-time an important factor for TA1? A17: The proposed computational hardware must be affordable under the RFMLS program. Additionally many algorithmic solutions may not be able to achieve real-time simply through adding additional processors. Q18: The TA2 P FA of 0.5 seems high. Can we propose solutions that do better? A18: Achieving the P FA could be difficult due to the small fraction of signals that are important, and the desired P D on important signals. The metric permits one unimportant signal for every two that you declare important. You may propose solutions that do better. Q19: Does the probability of detection metric given for TA2 apply both to anomalous as well as important signals? A19: Yes. Q20: How should TA1/TA2 & TA3 estimate level of effort associated with inter-ta interaction? A20: Proposers are expected to clearly identify the tasks and level of effort associated with inter-ta interaction. There is no restriction on TA1/TA2 and TA3 proposers discussing what the expected level of interaction is. Q21: Is there any preference on team size? A21: Proposers should propose what they believe is the appropriate team size and composition to address the tasks of this BAA. Q22: Is there a definition for technology transition given that this program calls for 6.2 research? A22: Examples of transition could include use of the technology in other research efforts as well as publication. Transition need not strictly refer to use in an operational DoD system.

4 Q23: Is publication about the TA3 hardware allowed? A23: Publication restrictions related to TA3 hardware will ultimately depend on the TA3 hardware selected. Q24: Slide 32 in the pdf of Industry Day slides could be interpreted as the power spectrum vs. time, for example, or I-Q data vs. time. Will the GFI data for Task 2A be I-Q data over time, similar to the GFI data for TA1? A24: All GFI datasets will be IQ data samples vs. time. Slide 32 did use a spectrum power density plot to illustrate the addition of important signals into backgrounds, but this does not represent the GFI data format. Q25: The BAA states in Task 1A that proposers should not plan to create or use their own signal samples. Is it permissible to augment the GFI signals data, which is a common deep learning practice? A25: The BAA states on page 11, Proposers should not plan to create or use their own signal samples. If proposers believe additional datasets are required they should explicitly state and justify the need to augment the GFI datasets. Any additional dataset generation to be performed under the proposed effort should be listed as a separate optional task. The BAA will be updated with this guidance. Q26: For task 1A, if an SDR is used to transmit different modulations, e.g. Bluetooth in one case and FM radio in another, are those considered different transmitters for the purposes of the Performance Metrics evaluation? A26: Task 1A datasets will consist of digital communications devices only. Each Task 1A dataset will focus on only a single signal type. The Task 1A metric does not apply across datasets of different signal types. The BAA will be updated with this information. Q27: Will the GFI I-Q data in Task 1A be for a single antenna, or will it be from a multi-antenna system like Fig 3 in the BAA? If the latter, will the Task 1A data include meta-data about the RF system settings? A27: GFI dataset recording hardware has not been identified. While it is desired to use multiple receivers, possibly multi-antenna receivers, proposers should not assume multiple simultaneous antennas are used. Metadata regarding the recording hardware may be provided. It is also desired to use multiple geometries, days, and weather conditions with associated metadata unlikely to be provided. Q28: Will Task 1A GFI data be digital modulations or both digital and analog modulations? A28: The proposed Task 1A solution should be robust to all common digital modulations.

5 Q29: In Task 1A, it is clear that the data will be in 0.1 sec snippets. What is the time span of GFI data samples in Task 2B? If it a longer continuous stream of data, then for the labelled important signals, do we need to identify the interval within the continuous stream of data that contains the important signal? Or do we consider the entire labelled sample to be important? A29: The data sample length for Task 1A has not been determined, but it is likely to be on the order of 1 second. The data samples length for Task 2A has not been determined but it is likely to be on the order of 10 minutes. Since important signals may last a relatively short time, the Task 2A system must identify the important signal temporal interval. Task 2B does not receive GFI datasets, but instead received realtime streaming data from the TA3 RF hardware. Since there is effectively no end to this dataset, identifying the temporal interval of the important signal is critical. Q30: For Task 1A, does the GFI dataset (I/Q samples) include samples from both Alice and Trudy transmitters talking at the same time? In other words, is there any interference from one transmitter to another? A30: No. Q31: For Task 1A, do GFI dataset I/Q samples represent down-converted baseband signal or direct output of an A/D converter? A31: Data recording and formatting has not been finalized, but baseband sampling is more likely. Q32: On page 19, the BAA states For planning and budgetary purposes TA2 performers should expect to deliver 5 RF systems. Is this meant to say TA3 performers? A32: Yes. TA3 (not TA2) proposers should budget for delivery of five RF systems. These should be complete systems including antennas. The RF systems delivered at 6 MAC must be the final systems with no upgrades planned for later delivery. Proposers may provide an option to upgrade the systems later in the program, but must provide justification as to the benefit to the Program of the optional upgrade. The BAA will be updated with this guidance. Q33: Will carrier frequency/phase recovery be implemented in the GFI data? A33: No. Q34: How are the hardware variations in the receiver removed from GFI data? A34: For Task 1A, please refer to question 27. A similar answer applies to Task 2A. There will be no attempt to remove variations in the data recording hardware.

6 Q35: Is there any guarantee that the Task 1A GFI contain enough features to differentiate 10,000+ devices? A35: Differentiation of 10,000+ devices appears plausible based on combination of published results using multiple different features. However, there is no guarantee that this is possible at all, or that the GFI dataset supports realization of this level of discrimination. Q36: How many Task 1A examples will be provided for each transmitter? The number of data available would have a direct consequence on the appropriate learning algorithm to use. A36: There will certainly be some transmitters for which there is only a single example. Other transmitters will have multiple examples (10-100), but may lack variation in environmental conditions across the samples. Q37: BAA (page 17) states, "For evaluation, the Task 2A system will be trained and tested using a previously unseen dataset and definition of importance". Please explain. A37: A Government laboratory will perform this testing and training following the process established by the Task 2A performer. The training data will have the same types of labels as the Task 2A GFI dataset. As is the case with the Task 2A GFI backgrounds, background signals will be unlabeled. The Government testing backgrounds will be similar but not identical to training backgrounds and GFI backgrounds. The definition of important will not change from training to testing, but it may be a definition that was not previously given as part of the GFI dataset. Q38: The Meeting and Travel Requirements section in page 48 of the BAA lists only the kick-off meeting, and the bi-annual PI meeting by the performers. The BAA also mentioned requiring quarterly technical review meetings on page 48. Is travel expected for the quarterly review meetings? A38: All key participants are required to attend the program kickoff meeting held at DARPA. Performers should also anticipate bi-annual program-wide PI Meetings held at DARPA, plus smaller (budget for two per year) off-cycle technical deep-dive meetings held at DARPA and periodic site visits at the DARPA Program Manager s discretion. For the quarters between bi-annual reviews, Government staff will travel to the performer facility for two quarterly reviews per year. A proposer with a geographically distributed team may need to budget some travel for these meetings. The BAA will be updated with this guidance. Q39: BAA page 19 middle of the third paragraph from the bottom reads, To support this the RF system provided by TA2 must have sufficient ability to exfiltrate raw, or near raw RF data through standard high bandwidth interfaces such as 10/40/100 Gb Ethernet, or other similar high-speed external buses. Should the reference to TA2 be TA3?

7 A39: Yes, TA3 (not TA2) is responsible for providing the RF systems and is also responsible for exfiltrating data to a nearby TA2 processor. The BAA will be updated with this correction. Q40: BAA page 20 Schedule & Deliverables reads; Integrate RF System to government dataset generation infrastructure. Will this government dataset generation infrastructure be open air or a cabled system? A40: To the extent possible all Government data is expected to be open air. This integration means that one or more of the five delivered TA3 systems may be used to create additional GFI dataset. The Government will provide the data recording equipment and solely operate the dataset generation process. Q41: Can we assume that for the Phase 3 open-air testing no motion is required (all HW is at fixed locations) so no NAV data will be needed or provided? A41: Yes. Q42: Will the Government or the TA3 provider pick the over-the-air test site? A42: The TA3 performer has full responsibility for the two outdoor real-time demonstration/tests, including selection of the location. The TA3 performer must identify and pay for the test location, and obtain access and authorization to test there. The BAA will be updated with this information. Q43: Can a TA3 vendor redistribute the total program funding to front load Phase 1 to better support HW build/test? A43: Proposers under all TAs should propose the funding schedule required to accomplish the proposed technical effort. TA3 is expected to be front-loaded for hardware procurement and end-loaded for demonstration and testing. Q44: Is the TA3 system required to transmit during the final two outdoor tests? A44: No. Q45: If a proposed TA3 RF system includes RF transmit functionality (in addition to the required RF receive functionality) for no additional cost, would that additional transmit feature favorably impact the proposal s evaluation? A45: No. But if selected for funding, a transmitting system may have a higher likelihood of finding post- DARPA transitions.

8 Q46: How many TA2 performers must TA3 support? Are the demonstrations concurrent or staggered? A46: This depends on the proposals received. TA3 proposers should estimate the number of TA2 performers and state the numbers of performers assumed for the TA3 pricing and schedule proposed. Since TA3 is responsible for all aspects of demonstration and testing, the TA3 proposer can choose either a concurrent or staggered schedule. The BAA will be updated with this guidance. Q47: Should TA1, TA2, and TA3 respondents break out Phase 2 and Phase 3 as separate cost options? A47: Yes Q48: Is there a requirement for a TA3 respondent to keep one or more of the completed RF systems at its own facility beyond 6 MAC? If so, will TA3 respondent need to provide outside parties with access to completed RF systems beyond 6 MAC? A48: TA3 proposers should budget for delivery of 5 complete RF systems by 6 MAC. Proposers may propose to keep one or more additional systems to aid with troubleshooting interface problems or planning for demonstration and testing. Proposers should not plan for providing external access to these additional internally held systems. The BAA will be updated with this guidance. Q49: The BAA indicates that inclusion of proprietary technology or information as part of the proposed work will negatively impact the likelihood of funding, and that proprietary information should not be included as part of the proposed work. Is this guidance limited to the new development solicited under Technical Areas (TA) 1 and 2, or is it intended also to apply to TA 3? Will TA 3 bidders offering SBIR technology/hardware be considered provided that the proposed offering does not preclude the Government's ability to transition the technology? A49: As stated during the Proposers Day, performers will have intellectual property ownership for all new development under the Program. The Government desires to retain at least Government-purpose rights. Asserted data rights restrictions will be assessed during proposal evaluation to assess the degree to which they may create barriers to technology transition. This applies to all TAs. Q50: Can the list of attendees for the 31 August 2017 Industry Day be made publicly available? A50: In order to protect personal privacy, DARPA will not release a list of Proposers Day attendees. Q51: In Task 1B, is the sensor node (Bob) the intended receiver of transmissions made by the transmitters that may have to be ACK d by Bob, or are the transmitting nodes talking to the other (non-bob) nodes? A51: In the notional scenario Bob is the intended communications recipient. However, for testing and demonstration another nearby communications system may be used to implement bi-directional

9 communications. The hardware implementation of Bob under Task 1B may be receive-only. The BAA will be updated with this information. Q52: In Task 1B, are there any limits on the type of feedback from the sensor node (Bob) or on how it is signaled, e.g., can feedback be signaled in a side channel such as Ethernet? A52: No. During the training phase Bob can provide feedback via any convenient channel. The BAA will be updated with this information. Q53: The Task 1A GFI appears to consist of a contiguous set of I/Q samples at 200Msps of some duration (called an RF sample ) and an associated label. For the purpose of scoping and costing the proposal, should proposers assume that each RF sample contains significant power from only one transmitter and therefore the ID tag indicates only 1 ID? A53: No. For example, signal sampling will not be synchronized with TDMA timing and samples may contain multiple sequential signals. Filtering may or may not be applied to isolate FDMA signals. CDMA signal samples may contain multiple overlapping signals. Other signals of not of interest may reside in the data. Q54: Regarding the TA2 Frequency plan: Does center frequency coverage mean that the center frequency is at any frequency between 1 and 6 GHz; or is the bandwidth from exactly from 1 to 6 GHz? A54: The frequencies available as input to Task 2B will depend on the TA3 systems proposed. For TA2 proposals, assume that a 500 MHz instantaneous bandwidth can be tuned anywhere in the range of 1-6 GHz. Q55: Are there any execution time or latency requirements for the TA2 algorithms to process data and control the TA3 hardware? A55: There is no explicit limit, and latency is not a problem for Task 2A. However, latency may degrade Task 2B performance against some fleeting signals. Q56: How many "open-air" demonstrations during Phase 3 (MAC 24-36) are required that the TA2 must performers support? How long is the planned duration for each open-air demonstration? A56: TA2 must support two formal TA3-run tests toward the end of Phase 3. TA3 will propose and plan all aspects of these tests, including the duration. TA2 must also support pre-test integration. The BAA will be updated with this information. Q57: The schedule on Page 8 shows that all TA2 software is complete and code is delivered at the end of Phase 2. However, during the Phase 3 demonstrations it may be beneficial to update the software

10 based on lesson learned from the open-air testing. Is this allowed? If so, are there any other considerations regarding software modifications during Phase 3? A57: During Phase 3, TA2 software changes may be made to fix problems discovered during Phase 2 or during pre-test integration. These changes must be limited to fixing capabilities implemented at the end of Phase 2. To minimize risks to a successful test result, no new capabilities may be added during Phase 3. The BAA will be updated with this information. Q58: The BAA requires an ACA between TA1 and TA3. However, the schedule on page 8 only shows that TA2 performers support TA3 open air demonstrations. With respect to the ACA, what technical support do the TA1 performers need to provide to TA3 during Phase 1 and Phase 2? A58: TA1 must be able to use input data received on the TA3 hardware system. Q59: The BAA states: Performers must provide their own RF hardware for Task 1B. In Phase 1, at least 10 nominally identical RF devices should be provided, for example, SDRs (software-defined radios). The RF systems must be commercially available so that a Government laboratory can purchase identical units for verification testing. Alternatively, additional RF systems can be provided by Task 1B performers to the government team for independent evaluation. Does this mean that performers must internally procure and provide their own RF hardware, and that the hardware costs for the RF devices cannot be part of the proposal? Or, does it simply mean that the RF devices will not be GFE, and performers can propose hardware costs for the RF devices? A59: This simply means that the hardware is not GFE. Proposals may include the cost of purchasing or renting required hardware. Proposers may also propose to use hardware available at no cost to the Program. Proposers should review the cost-related proposal evaluation criteria published in the BAA and also consider that DARPA has a limited budget. Q60: Under Task 1B, the cost of 100 SDRs for ac or n would be significant if MIMO support is required. Do you want to support MIMO data links in TA 1B? What instantaneous bandwidth do you want to support with the transmitters for Task 1B? If we are to support ac transmit, the cost for SDRs will jump considerably, due to the IBW. A60: Proposers may propose any standard protocol and any SDR for Task 1B. There is no required protocol or SDR bandwidth. Proposers should review the cost-related proposal evaluation criteria published in the BAA and also consider that DARPA has a limited budget. Q61: The BAA states TA1 and TA2 proposers should provide separate options for each of the two capability tasks. Does this mean that we should propose separate options for 1A (RF Feature Learning) and 1B (Waveform Synthesis)? A61: Yes. Tasks 1A and 1B should be separately costed, as should Tasks 2A and 2B. This provides the Government the option to fund only a single Task under a TA.

11 Q62: Can multiple proposals be submitted from the same organization (same CAGE code) for TAs 1 and/or 2? A62: There is no limit on the number of proposals that a single organization may submit. Q63: For Task 1A, if a new transmitter appears multiple times, do you expect the system to get the same "new" label each time? If more than one new transmitter appears, do you expect the system to automatically create different "new" labels, and to recognize repetitions? A63: While the emphasis is offline learning, the Task 1A system must be able to recognize that a transmitter is new. The TA1 system must assign an ID to that new transmitter and then produce the same ID when the transmitter is seen again in the future. The BAA will be updated with this clarification. Q64: What information is shared between Alice and Bob for Task 1A and Task 1B? Do you expect that Alice will know Bob's model, or does Alice have to infer her own uniqueness? A64: Under Task 1A, no information is shared. Under Task 1B, Bob shares information on how well Alice can be distinguished from Trudy and from transmitters in the Task 1A database. Alice does not need to infer her own uniqueness. Q65: For Task 2A, the metric is for attention, not saliency. Do you have a saliency metric in mind? A65: The saliency test is realized by changing the definition of important to be some specific or generally anomalous signals. Q66: Could the protocol/waveform used in the TA2 training or evaluation dataset include an LPI waveform? A66: Yes, but if it is included that will not be disclosed. Q67: Page 33 of the BAA, from Section I, Cover Sheet, Item (12), states the total proposed cost are to be provided by calendar year and by government fiscal year (GFY), but in Section II. Detailed Cost Information (Prime and Subcontractors) states that the total cost provided in the Summary Cost Breakdown should be organized by phase and performer s fiscal year. Could the Government please clarify that this is the intent? A67: Please follow the guidance of Section II and provide a cost summary by phase and performer fiscal year. The BAA will be amended to reflect this guidance. Q68: Can TA1 phase 1 and TA1 phase 2 be performed concurrently, or must they be performed sequentially?

12 A68: They must be performed sequentially. Q69: The BAA stresses that the proposal must precisely adhere to the BAA formatting instructions. The BAA format instructions do not clearly specify whether sub-sections are allowed within Technical/Management volume, Section II.A, Technical Approach. Since the technical approach may contain sub-tasks/sub-sections, would the use of sub-sections within a proposal when addressing II.A be permissible? A69: Sub-sections may be used as necessary. Q70: Is it correct to assume that the updated TA1 dataset gathered using TA3 hardware will be formatted to be similar or the same as the original dataset? A70: Yes, with small variations to be expected such as the sample rate of the data, bit-depth of each sample, etc. Q71: Are there any labels other than important and unimportant provided for the TA2 dataset? A71: Additional labels which describe important signals may be provided, such as signal name or modulation. Proposed approaches should be able to reconcile these additional labels. Q72: Might the TA2 GFI unimportant background sometimes have an important signal in it? A72: Yes, but not intentionally. Q73: Is important always a single set of labels? A73: The important signals in the training database are exemplars from the class of important signals. The TA2 solutions should not be looking for a set of labels exactly identical to the training signals. Q74: What is the definition of a detection for the TA2 PP DD performance metric? (Waveform, part of a waveform, sequence of events, ) A74: The TA2 system reports a set of labels for each signal deemed important. The desired labels will be defined in the training dataset which has not yet been created. Correlation between the reported labels and important signal truth data will determine whether the report is graded as a detection or a false alarm. Perfect correlation will not be required for the report to be declared a detection. A more precise mathematical definition of detection will be provided with Task 2A GFI datasets. Q75: How will the government team evaluate Task 2B solutions?

13 A75: For Task 2A, a Government laboratory will apply performer algorithms to new training and evaluation datasets. For Task 2B, a Government laboratory will apply performer algorithms to one of the RF systems delivered by TA3. Q76: Does the TA3 RF system have to provide processing for TA2 algorithms? A76: TA2 algorithms will be hosted on an external compute platform provided by TA2. TA3 systems are not required to provide any additional computational capabilities to host TA2 algorithms. However any computational elements resident on the TA3 system may be used at the joint discretion of TA2/TA3 performers. Q77: Are TA2 and TA3 responsible for post demonstration analysis and reporting? A77: Yes. Q78: The BAA states that the TA3 system is required to provide up to 500MHz of instantaneous bandwidth (IBW). Can this performance metric be met by using multiple simultaneous receive chains which in aggregate are equal or greater than 500 MHz? A78: The government desires each individual receive chain be capable of up to 500MHz IBW. Q79: The BAA states that Task 1A must be able to stream data. Does this require a real-time RF sensor be integrated with Task 1A, or is streaming playback of previously gathered data acceptable? A79: Streaming playback of previously gathered data is acceptable to meet the real-time requirement. Q80: Is custom compute hardware acceptable to use for TA1 and TA2? A80: The RFMLS program does not expect to develop new digital compute hardware, however proposers may elect to use already existing custom compute platforms. Should custom compute platforms be proposed, performers should schedule and budget to deliver a system to government at months 11 and 23 for evaluation. The BAA will be updated with this guidance. Q81: Should TA1 / TA2 proposers expect to deliver both training and evaluation and evaluation software and necessary documentation? If so is it acceptable that during training the system augments some initial parameter set? A81: Yes proposers should expect that part of government evaluation will include retraining their system. Training may start from a performer defined default state.

14 Q82: Should the TA3 system be capable of providing multiple beams simultaneously? A82: This is permissible under the BAA. Q83: Should TA1 and TA2 proposals include an option for each Task for each Phase? A83: TA1 and TA2 should propose a base consisting of both Tasks. Options should be proposed for follow on phases. In both the base and option(s), Tasks A and B should be proposed as separate tasks. Q84: Per the BAA, p. 10, which states evaluation conducted by the government team will include evaluation against a previously undisclosed communications waveform will government evaluators train the TA1 system on examples of the new communications waveform prior to evaluation? A84: Yes multiple examples of multiple transmitters using the new communications waveform will be provided to the system by evaluators during training, prior to evaluation. Q85: The BAA states on p. 11, after training, the evaluation dataset will contain new examples from many of the transmitters in the training database, plus signals from new transmitters will the new transmitters referred to here, be of the same type as those in the training set? A85: Yes, the new transmitters are new identities using the same type of communications waveform/protocol as the other examples in the training set. Q86: While training the Task 1B system, is it acceptable to assume a secure environment, i.e., that the true identities of all devices are known? A86: Yes. Q87: Does the list of references count against the 25 page limit for Technical Volume? A87: No. Q88: Do the PP DD metrics given for TA1 and TA2 include identifying the proper label? A88: Yes. Q89: May a Task 1B approach insert a device in the RF-chain of an existing ASIC-based communications standard? A89: Yes.

15 Q90: Should co-channel interference be expected in TA1 and TA2 datasets? A90: To the extent possible TA1 and TA2 datasets will generated in an in-the-wild setting, and should be expected to contain all manner of impairments one might expect in real-world setting, including, but not limited to: co-channel interference, multi-path, doppler, simultaneous near and far signals, etc. Q91: How many transmitters are expected in the TA1 dataset? A91: Greater than 1000 transmitters are expected in the TA1 dataset. Q92: For SBIR subcontractors are IP data rights considered government rights for the purposes of this BAA? A92: None of the work being conducted under the resulting contracts will, itself, be considered SBIR research. However, any preexisting data being delivered under a resulting contract that was developed/delivered under a previous SBIR contract should be listed in the data rights assertions list/certification as carrying SBIR rights per DFARS Part 227. Q93: For TA3 will DARPA consider a Firm Fixed Price, Commercial Item award instrument? A93: No - DARPA is not seeking an already existing commercial product/service. RFMLS is a Science and Technology (S&T) program. A firm fixed price research contract can, however, be used for the conduct of S&T if, for example, the performer does not possess an approved cost accounting system (but such an instrument will not be awarded under FAR Part 12 as a commercial Item procurement). In such proposals, the performer should include a payment milestone plan. Q94: For TA3 will DARPA consider different award instruments for different Phases? (e.g. Firm Fixed Price for Phase 1, Cost-Plus for Phase 2 and 3?) A94: Yes. Q95: Are there are any restrictions to using a foreign university as one of our subcontractor? A95: No. Q96: We intend to bid a traditional CPFF contract as a prime. Our bid will likely include a commercial vendor as a subcontractor who has not done any business with the government in the last several years. What FAR/DFAR rules will likely be flown down to them, if our proposal is selected? A96: The proposer needs to consider what role this commercial vendor will have on the project. If they will simply provide a standard commercial service (e.g., not be a team member) then they are likely a

16 vendor and would receive from the prime a purchase order having any flowdowns required by the FAR/DFARS (which the prime needs to research and understand). If the vendor will be serving as a team member working to directly contribute to the tasks delineated in the prime's SOW then they are likely a full subcontractor with all of the required FAR/DFARS flowdowns. In either case, it is the prime contractor's responsibility to understand what FAR and DFARS clauses will be required in the chosen award instrument. New responses added 10/3/2017 Q97: Is there a required/suggested excel spreadsheet template that we should use to prepare the cost proposal? A97: No. While the BAA states that Proposers are required to provide the aforementioned cost breakdown as an editable MS Excel spreadsheet, there is no required or suggested template. Please read the BAA section titled Section II. Detailed Cost Information (Prime and Subcontractors) thoroughly to ensure that you are including all of the required content. Q98: What are the polarizations of the transmit and receive antennas in the GFI datasets? A98: A complete answer cannot be provided since the GFI datasets have not yet been created. Nearly all of the transmit antennas will be outside the control and cognizance of the experimental personnel. Thus, the transmit polarizations will be unknown and uncontrolled. The initial GFI datasets will likely use a single linear receive polarization with orientation not yet selected. A limited dual-linear polarized receive dataset is under consideration. Later in the Program when TA3 hardware is used to create additional GFI datasets, the receive polarization will be determined by the TA3 hardware capability. The BAA does not require any specific TA3 polarization and the ultimate TA3 receive hardware polarization is currently unknown. Q99: TA1B Phase 2, according to the industry day slides, expects an evaluation and demo capability on 100 SDRs. Does this mean that TA1 performers are expected to cost, bid and build a hardware testing rig that contains 100 SDRs, which may cost anywhere from $1-5k apiece, plus rack infrastructure, power, cooling, etc.? A99: The BAA states, In Phase 1, at least 10 nominally identical RF devices should be provided, for example, SDRs (software-defined radios). During Phase 2, Task 1B performers will improve the performance of their system and expand the size of the overall population of transmit devices used to at least 100. Larger populations are desirable to the extent practical. Yes, proposers may select SDRs to serve as their transmit devices and the cost of these devices should included in the proposal, at least 10 under Phase 1 and the remainder under the Phase 2 option. Also see the answer to Q60 which states: Proposers should review the cost-related proposal evaluation criteria published in the BAA and also consider that DARPA has a limited budget.

17 Q100: On page 30 of the proposal within section F (Teaming), what is meant by the term team member as it relates to level of participation formally committed? Is DARPA looking for Key Personnel or clarification of the commitment of each organization (i.e. does team member refer to a person or an entity)? A100: Page 30 of the BAA states, Describe the teaming agreements proposed, including all subcontractors and consultants. Detail the level of participation formally committed by each team member, and clearly identify details of the teaming plan that have not yet been finalized. The definition of team member is subcontractors and consultants and refers to organizations, where some consultants may be a single-person organization. The emphasis is on formal organizational commitments and is not a request for identification of key personnel. Q101: On page 26 of the BAA it states All pages shall be formatted for 8-1/2 by 11 inch paper with type not smaller than 12 point. We have noticed in many recent BAAs that smaller fonts are acceptable in tables and figures (and captions). Does the preceding quote from page 26 apply to all text including tables and figures (and captions)? A101: The purpose of the proposal formatting guidance is to ensure that reviewers reading printed proposals will be able to read all of the proposal. Proposals will not be summarily rejected if smaller fonts are occasionally used. However, if reviewers are unable to read small-font portions of the proposal then the overall appreciation of the proposal may suffer and the specific technical content of these small-font portions will not contribute toward the reviewers assessment of the proposal. Q102: Please clarify the proposal due date. May we receive an extension beyond the proposal due date of 1 PM 10/10/2017? A102: There will be no change to the submission schedule. It is not possible for DARPA to grant an extension to one proposer without granting the same extension to all proposers. Note that the BAA is already structured to receive proposals after the initial due date. The BAA states: The full proposal must be submitted via the DARPA BAA website on or before 1:00 p.m., EST 10 October 2017 in order to be considered during the initial round of selections; however, proposals received after this deadline may be received and evaluated up to five months (150 days) from date of posting on FedBizOpps. Full proposals submitted after the due date specified in the BAA or due date otherwise specified by DARPA after review of proposal abstracts may be selected contingent upon the availability of funds. Proposers are warned that the likelihood of available funding is greatly reduced for proposals submitted after the initial closing date deadline. Failure to comply with the submission procedures may result in the submission not being evaluated.