Development of real-time Deep Reinforcement Learning (DRL) Hardware

Organization DEVCOM Army Research Laboratory... Reference Code ARL-C-CISD-300144 Description About the Research Current approaches optimize machine learning training largely by exploiting Deep Neural Networks (DNNs) sparsity. The compute-intensive floating-point 32-bit representation represents remaining non-zero valued network parameters. These approaches need to be improved to develop a real-time hardware implementing Deep Reinforcement Learning algorithms for the tactical arena. Additionally, High Level Synthesis (HLS) will be incorporated to obtain hardware designs optimized for various criteria, e.g. power, latency, computation, etc. ARL Advisor: Vinod K. Mishra ARL Advisor Email: vinod.k.mishra.civ@army.mil About CISD The Computational and Information Sciences Directorate (CISD) conducts research in a variety of disciplines relevant to achieving and implementing the so-called digital battlefield. Problems address the sensing, distribution, analysis, and display of information in the modern battle space. CISD research focuses on four major areas: communications, atmospheric modeling, battlefield visualization, and computing About ARL-RAP The Army Research Laboratory Research Associateship Program (ARL-RAP) is designed to significantly increase the involvement of creative and highly trained scientists and engineers from academia and industry in scientific and technical areas of interest and relevance to the Army. Scientists and Engineers at the CCDC Army Research Laboratory (ARL) help shape and execute the Army's program for meeting the challenge of developing technologies that will support Army forces in meeting future operational needs by pursuing scientific research and technological developments in diverse fields such as: applied mathematics, atmospheric characterization, simulation and human modeling, digital/optical signal processing, nanotechnology, material science and technology, multifunctional technology, combustion processes, propulsion and flight physics, communication and networking, and computational and information sciences. A complete application includes: • Curriculum Vitae or Resume • Three References Forms • An email with a link to the reference form will be available in Zintellect to the applicant upon completion of the on-line application. Please send this email to persons you have selected to complete a reference. • References should be from persons familiar with your educational and professional qualifications (include your thesis or dissertation advisor, if applicable) • Transcripts • Transcript verifying receipt of degree must be submitted with the application. Student/unofficial copy is acceptable If selected by an advisor the participant will also be required to write a research proposal to submit to the ARL-RAP review panel for : • Research topic should relate to a specific opportunity at ARL (see Research Areas) • The objective of the research topic should be clear and have a defined outcome • Explain the direction you plan to pursue • Include expected period for completing the study • Include a brief background such as preparation and motivation for the research • References of published efforts may be used to improve the proposal A link to upload the proposal will be provided to the applicant once the advisor has made their selection. Questions about this opportunity? Please email ARLFellowship@orau.org. Eligibility Requirements">11 ) • Degree: Any degree . • Academic Level(s): Any academic level. • Discipline(s): • Computer, Information, and Data Sciences