EndGen_TA1_RFI_C.pdf

Request for Information (RFI) on Endless Generative (End-Gen) Waveforms for Communications

Agency: Office of the Director of National Intelligence Office: Intelligence Advanced Research Projects Activity

RFI Number: IARPA-RFI-24-06

Synopsis:

The Intelligence Advanced Research Projects Activity (IARPA) seeks information regarding methodologies capable of arbitrary generation of communication waveforms in a wide variety of propagation media. Point-to-point communications are at the heart of innumerable mission spaces spanning a host of different objectives. At present, most communications channels make use of fixed waveforms designed to meet the demands of the “average” propagation media and the available communication equipment. The static nature of these waveforms can be disadvantageous to the stability of the communication channel given spatial-temporal variations in the propagation media. Therefore, communications frameworks that can generate on-demand waveforms are especially attractive when the current time and cost of developing new waveforms (1), to meet fluid mission needs, is prohibitively high.

This RFI seeks methodologies that, when coupled with appropriate communications hardware, show the potential to achieve on-demand waveforms that meet the required human objectives and limitations of the propagation medium. Such an approach could result in endless generative (End- Gen) waveforms that automatically optimize the communications channel without the constraints of human-generated communications protocols. End-Gen nodes, therefore, will contain frameworks which produce generative waveforms that satisfy any generic mission objective function.

End-Gen nodes that create End-Gen waveforms are of keen interest as they allow for the rapid deployment of waveforms to match the real-time change in mission needs. Specifically, IARPA is interested in demonstrating three parts of any End-Gen paradigm: 1) the processing framework needed to generate meaningful communications links, 2) the hardware SWaP (size, weight, and power) required to run the framework, and 3) the sensors/transducers needed to couple framework outputs to the propagation medium.

This RFI is for planning purposes only and does not constitute a formal solicitation of proposals to fund an End-Gen program for arbitrary waveform synthesis. The RFI is particularly interested in respondents who have experience with breadboard/brassboard-level lab demonstrations to prove feasibility of generative waveform frameworks. While analytical or simulated responses to the RFI questions below will be accepted, any response should cast the solution into how your analytical or simulated results could be used in a laboratory experiment. Additionally, while artificial intelligence and machine learning (AI/ML) methods are commonly expected, non-AI/ML methods are also welcome; however, any submitted method should generate waveforms on-the-fly and not resort to table lookups of preexisting waveforms.

Background:

Numerous survey papers (2) (3) (4) have discussed the role that nontraditional approaches to communications could take in all layers of next-generation protocol stacks. These roles include anything from radios that learn to optimize medium access control (MAC) between multiple nodes, to AI/ML-learned error correction codes, channel filtering, and modulation at the physical (PHY) layer. While these are all interesting uses of nontraditional comms development, this RFI is focused specifically on solutions for the observable PHY layer waveform itself; in other words, bits are input into your End-Gen framework and the waveform that results is based on some mission criteria/objective. In general, frameworks for in situ approaches to communications are hampered by several key problems such as the lack of feedback channels and the processing required to sufficiently train smart radios. Since these issues with simulation models can obscure unrealizable solutions, this RFI is particularly interested in respondents that can show end-to-end demonstrations of generative waveform frameworks. For example, Deepmod (5) details an experiment that uses deep learning AI/ML models for over-the-air training. In this framework, backpropagation is addressed by creating local adversarial critics, and the entire framework has been shown to run on radio frequency (RF), powerline, and acoustic channels with standard desktop processors connected to software radios.

Scope:

This RFI is scoped for frameworks that can generate an arbitrary number of PHY layer waveforms.

All frameworks need to accept information-bearing bits into the framework, appropriately consume human objectives, ingest observations of the channel, and then generate appropriate analog outputs. These outputs must then be transduced onto the wired, or wireless, media. All types of channel media are of interest (e.g., powerline channels, free space RF, optics, acoustic/ultrasonic, etc.). The envisioned models and hardware could enable rapid pivots in mission spaces as human-developed waveforms would no longer delay the specific mission needs.

This RFI is interested in any smart radio that couple objective functions with processing frameworks to communicate in any media with appropriate energy propagation.

Specifically, respondents capable of designing, or running, experiments demonstrating generative waveforms are of interest. Consider the ubiquitous point-to-point communications channel shown:

For lab experiments, such a setup could be accomplished using a digital-to-analog converter (e.g., software-defined radios (SDR), RF system-on-chip (RFSoC), or low-rate sound cards). A typical experiment would, at a minimum, follow the steps:

A. During learning stages, End-Gen Node A and End-Gen Node B exchange energy across the noisy channel cognizant of human-provided objective(s), and channel observations, to create an information-bearing waveform.

B. The waveform objective should be human-understandable and could be anything, from “Be bursty,” to “Maximize SNR for the given bandwidth,” to “Match a specific spectral width,” etc. The human objective must map to symbolism understood by the framework.

C. After learning, End-Gen Node A and End-Gen Node B communicate across the physical medium with the final waveform, which waveform ideally meets the objective(s).

Any change in the channel, objective, and/or other environmental effects should also change the final waveform hence End-Gen waveforms are arbitrarily numerous.

Request:

This RFI seeks a better understanding of laboratories capable of designing, creating, and testing endless generative waveforms. Such experiments should be able to handle a wide variety of channel media with appropriate transducers and have a suite of selectable objective functions. It is also of interest to understand how one can score the performance of a framework’s output waveform against a prototypical human-generated waveform in a similar environment. Questions for respondents are broken down into five different categories.

Questions for Respondents

1. Objectives: The criteria, as required by the human for a specific channel, that the End-Gen waveform is trying to meet.

a. What common comms objectives such as “minimize bit-error rate” or “meet throughput requirements” should be in all End-Gen nodes?

b. Other objectives, such as “match a specific spectral width” or “minimize peak power” or “be frequency agile”, are only limited by human imagination. What other objectives might help create endless waveforms in meaningful ways?

c. How are all objective functions understood mathematically by the End-Gen node?

2. Framework: Any model that takes input bits, Objectives, and observations of the channel to output in the form of some transducible quantity (e.g., voltage).

a. What methodologies are you suggesting to “train” End-Gen nodes? How does the methodology improve communication quality over time (e.g., how does it “learn” to improve the quality of the communication channel)? What are the convergence criteria and times?

b. How is backpropagation, or equivalent weight updating mechanism, achieved when no side channel is available?

c. How are channel observations, or other sensor inputs, used for improving (i.e., learning)?

3. Processor: The hardware (HW) used to run the framework and interfaces between it and the hardware transducers.

a. How does the size of the End-Gen framework influence hardware SWaP and performance? Can the framework scale down in HW?

b. What kind of HW, such as RFSoC, SDR, or field-programmable gate arrays (FPGA) might be used in an experiment?

c. What trade-offs in HW are needed for different propagation media?

4. Transducer: At the transmitter node, the transducer is simply the HW needed to convert framework outputs to appropriate energy types needed by the communication environment.

Conversely, at the receiver node, the transducer converts energy back into whatever format is needed for the node to consume.

a. How is the output waveform converted to energy appropriate for the medium and vice versa?

b. Can the same frameworks be used for entirely different propagation media by swapping transducers?

5. Metrics: The performance of End-Gen frameworks must be quantifiable as it relates to the human-generated Objective and basic communications requirements.

a. What metric score(s) can be provided to determine success? Success is defined by how close the End-Gen waveform meets the Objectives.

b. How can the End-Gen waveform be compared to offline human-generated waveforms?

Respondents who have not performed experiments, but who have valid simulated approaches, should describe in detail how the experiments might be performed in the lab.

Preparation Instructions to Respondents:

IARPA requests that respondents submit responses to the above questions for use by the Government in formulating a potential R&D program. IARPA requests that submittals briefly and clearly describe: (i) how approaches to lab experimentation could be run for generative waveforms,

(ii) the training datasets to use in the approach, (iii) how the approach may be quantified against conventional approaches, and (iv) comment on the performance of the approach. If appropriate, respondents may also choose to provide a non-proprietary rough order-of-magnitude (ROM) estimate of what such approaches might require in terms of funding and other resources to execute the respondent’s vision for a framework of endless generative waveforms in a timeframe of four years or less. This announcement contains all the information required to submit a response. No additional forms, kits, or other materials are needed.

IARPA welcomes responses from all capable and qualified sources from within and outside of the U.S.

Reponses must meet the following formatting requirements:

1. A one-page cover sheet that identifies the title, organization(s), respondent's technical and administrative points of contact - including names, addresses, phone numbers, and email addresses of all co-authors, and clearly indicating its association with IARPA-RFI-24-06;

2. A substantive, focused, one-half page executive summary;

3. A description of the technical challenges and suggested approaches to experiment with End-Gen waveforms in a variety of propagation media. Responses should answer most of the suggested questions in each of the five categories of: Objectives, Framework, Processor, Transducer, and Metrics and be limited to 3 pages total (excluding cover pages, executive summary, and citations) in minimum 12-point Times New Roman font, appropriate for single-sided, single-spaced 8.5 by 11-inch paper, with 1-inch margins.

Please provide diagrams, figures, charts to help illustrate concepts when possible; and

4. A list of citations (any significant claims or reports of success must be accompanied by citations) especially citations that reflect published work for experiments like End-Gen waveforms. There is no page limit for citations.

Submission Instructions to Respondents:

Responses to this RFI are due no later than 5:00 p.m., Eastern Time, June 23, 2024. All submissions must be electronically submitted to dni-iarpa-rfi-24-06@iarpa.gov as a PDF document. Inquiries to this RFI must be submitted to dni-iarpa-rfi-24-06@iarpa.gov. Do not send questions with proprietary content. No telephone inquiries will be accepted. In response to submissions to this RFI, IARPA may organize an invitation-only workshop for the purpose of reviewing and discussing relevant current and future research. Should a workshop occur, participants may be asked to make formal presentations based on their RFI submissions or associated research. Written submissions and information discussed at a workshop may assist in the formulation of future U.S. Government research.

Disclaimers and Important Notes:

This is an RFI issued solely for information and planning purposes and does not constitute a solicitation or authority to enter into negotiations for a contract. Respondents are advised that IARPA is under no obligation to acknowledge receipt of the information or to provide feedback to respondents with respect to any information submitted under this RFI. Responses to this notice are not offers and cannot be accepted by the Government to form a binding contract. Respondents are solely responsible for all expenses associated with responding to this RFI. IARPA will not provide reimbursement for costs incurred in responding to this RFI. It is the respondent's responsibility to ensure that the submitted material has been approved for public release by the information owner.

The Government does not intend to award a contract on the basis of this RFI or to otherwise pay for the information solicited, nor is the Government obligated to issue a solicitation based on responses received. No proprietary and no classified concepts or information shall be included in the submittal. However, should a respondent wish to submit classified concepts or information, prior coordination must be made with the IARPA Program Security Officer. Email the Primary Point of Contact with a request for coordination with the IARPA Program Security Officer. Input on technical aspects of the responses may be solicited by IARPA from non-Government consultants/experts who are bound by appropriate non-disclosure requirements.

Contracting Office Address:

Office of the Director of National Intelligence Intelligence Advanced Research Projects Activity Washington, District of Columbia 20511 United States

Primary Point of Contact:

Dr. Adam L. Anderson, Program Manager dni-iarpa-rfi-24-06@iarpa.gov Intelligence Advanced Research Projects Activity Sources:

1. 6G Wireless Communications Networks: A Comprehensive Survey. Alsabah, M. 2021, IEEE Access, vol.

9, pp. 148191-148243.

2. 6G White Paper on Machine Learning in Wireless Communication Networks. Ali, Samad, et al. 4 2020.

3. A Survey on Machine Learning-Based Performance Improvement of Wireless Networks: PHY, MAC and Network Layer. Kulin, Merima, et al. 3, 1 2021, Electronics, Vol. 10, p. 318. 2079-9292.

4. Machine Learning in the Air. Gunduz, Deniz, et al. 4 2019.

5. Deep modulation (Deepmod): A self-taught PHY layer for resilient digital communications. Anderson, Adam, et al. 2019, arXiv preprint arXiv:1908.11218.