You are here
Utilizing ML Algorithms to Track and Identify UAS Threats
Phone: (256) 562-0087
Phone: (256) 562-0087
Frequency-modulated continuous wave (FMCW) lidar is the optimal solution for cost-appropriate ground-based imaging and discrimination of small UAS at the 3 km range. In this effort Polaris will first design and assess various FMCW lidar system configurations for SOCOM’s use case. Second, Polaris will employ an innovative approach to machine learning (ML) in which innovative techniques which offer transparent attribution, are robust to noise, and use relatively modest amounts of training data to achieve desired results are used to conduct discrimination and identification of UAS platforms. Third, atmospheric turbulence effects will be corrected in the FMCW lidar data using innovative techniques developed by Polaris personnel. Turbulence correction is critical to achieving the resolution required to image Group 1 UAS (~ 1 cm transverse) at extreme ranges. At the completion of Phase I, the SOCOM will receive an FMCW lidar system design optimized for specific use case and mission constraints. This design will be complete with a parts list, cost, and schedule for joint construction of the system with our partners. Additionally, SOCOM will receive a detailed roadmap describing the ML solutions which will be developed by Polaris in Phase II of this program and beyond. The roadmap will focus on the near-term delivery of workable discrimination and identification (D&I) ML algorithms. These near-term solutions will be followed by a transition into the development of an increasingly robust D&I ML capability, specifically tailored to the delivered FMCW lidar systems, as well as a focus on supporting algorithms currently in use. Polaris’ final deliverable at the end of Phase I will be an assessment of the optimal approach for integrating our still-frame turbulence mitigation algorithm into an enhanced lidar full motion video (FMV) turbulence mitigation capability. This approach to real time turbulence mitigation will weigh the efficacy, speed, and power usage of the proposed real-time algorithm, as well as cost of development and time required for implementation.
* Information listed above is at the time of submission. *