Autonomy, Army STTR, Phase I

Optical Computing Network

Release Date: 04/19/2023
Solicitation: 23.B
Open Date: 05/17/2023
Topic Number: A23B-T003
Application Due Date: 06/14/2023
Duration: Up to 6 months
Close Date: 06/14/2023
Amount Up To: $197,000

Objective

Design and build a programmable optical network equivalent to an electrical network to solve Markovian graphs with cycles. Forney-style factor graphs can be solved while avoiding the creation of trees.

Description

The use of digital image processing to enable target detection, classification, recognition and identification, as well as targe state estimation for fire control solutions is computationally intensive. It requires significant processing power, which in turn requires significant electrical power.

A programmable optical network can be used to perform these computations at reduced Size weight and power and at faster speeds. Factor graphs have been used to describe Bayesian networks (Pearl, 1988) and were applied to SLAM (Simultaneous Location and Mapping) by Dellaert (2017). These problems tend to be decomposed into trees for solution.

The most general graph, and the one that is most difficult to solve, is the undirected graph with cycles. This is related to quantum computing and such difficult logistical problems such as the travel salesman conundrum.

It is desirable to try to develop a room temperature solution, based on optical networks, that can at least reliably solve all convex Kalman filter problems.

Phase I

Design and develop programmable optical circuit elements that map the nodes in a factor graph to those of an optical network much as Vontobel did for electrical components.

Phase II

Develop and demonstrate a prototype system consisting of the optical elements to create a network that can solve a problem.

Phase III

Build an integrated optic that can be deployed that can implement a Kalman filter with real world application.

Submission Information

Please refer to the 23.B BAA for more information. Proposals must be submitted via the DoD Submission site at https://www.dodsbirsttr.mil/submissions/login

STTR Topic

References:

  1. Dellaert, F. (2017). Factor Graphs for Robot Perception; Foundations and Trends in Robotics, Vol 6. No. 1-2 (2017) 1-139; DOI: 10.1561/2300000043
  2. Pearl, J. (1988), Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference; Morgan Kaufmann Publishers Inc. San Francisco Calf; ISBN 1-55860-479-0
  3. Vontobel, P.O.; Factor Graphs, Electrical Networks, and Entropy; http://www.isiweb.ee.ethz.ch/papers/arch/pvto-dlip-aloe-2002-mtns.pdf
  4. Vontobel, P.O.; Kalman Filtering, Factor Graphs and Electrical Networks. http://www.isiweb.ee.ethz.ch/papers/arch/pvto-dlip-aloe-2002-mtns.pdf
  5. Wang,S. (2020); A Factor Graph-Based Distributed Consensus Kalman Filter; IEEE Signal Processing Letters, Vol 27, 2020.

Objective

Design and build a programmable optical network equivalent to an electrical network to solve Markovian graphs with cycles. Forney-style factor graphs can be solved while avoiding the creation of trees.

Description

The use of digital image processing to enable target detection, classification, recognition and identification, as well as targe state estimation for fire control solutions is computationally intensive. It requires significant processing power, which in turn requires significant electrical power.

A programmable optical network can be used to perform these computations at reduced Size weight and power and at faster speeds. Factor graphs have been used to describe Bayesian networks (Pearl, 1988) and were applied to SLAM (Simultaneous Location and Mapping) by Dellaert (2017). These problems tend to be decomposed into trees for solution.

The most general graph, and the one that is most difficult to solve, is the undirected graph with cycles. This is related to quantum computing and such difficult logistical problems such as the travel salesman conundrum.

It is desirable to try to develop a room temperature solution, based on optical networks, that can at least reliably solve all convex Kalman filter problems.

Phase I

Design and develop programmable optical circuit elements that map the nodes in a factor graph to those of an optical network much as Vontobel did for electrical components.

Phase II

Develop and demonstrate a prototype system consisting of the optical elements to create a network that can solve a problem.

Phase III

Build an integrated optic that can be deployed that can implement a Kalman filter with real world application.

Submission Information

Please refer to the 23.B BAA for more information. Proposals must be submitted via the DoD Submission site at https://www.dodsbirsttr.mil/submissions/login

References:

  1. Dellaert, F. (2017). Factor Graphs for Robot Perception; Foundations and Trends in Robotics, Vol 6. No. 1-2 (2017) 1-139; DOI: 10.1561/2300000043
  2. Pearl, J. (1988), Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference; Morgan Kaufmann Publishers Inc. San Francisco Calf; ISBN 1-55860-479-0
  3. Vontobel, P.O.; Factor Graphs, Electrical Networks, and Entropy; http://www.isiweb.ee.ethz.ch/papers/arch/pvto-dlip-aloe-2002-mtns.pdf
  4. Vontobel, P.O.; Kalman Filtering, Factor Graphs and Electrical Networks. http://www.isiweb.ee.ethz.ch/papers/arch/pvto-dlip-aloe-2002-mtns.pdf
  5. Wang,S. (2020); A Factor Graph-Based Distributed Consensus Kalman Filter; IEEE Signal Processing Letters, Vol 27, 2020.

STTR Topic

Optical Computing Network

Scroll to Top