EpiSci

Team name

Team EpiSys Science (EpiSci) 

Awards

• UAS 5.0: First Responder UAS 3D Mapping Challenge
  • Stage 4 – Winner: $30,000
  • Total 5.0 Prizes: $65,000
• UAS 3.2: LifeLink 
  • 3rd Place: $10,000
  • Total 3.2 Prizes: $25,000

UAS 5.0: 3D mapping solution summary

Team EpiSci has developed a high-performance 3D mapping software solution for a class of small-sized, lightweight, and highly affordable UAV platforms. Leveraging proprietary computer vision and sensor fusion algorithms, our solution achieves two primary objectives: real-time construction of high-fidelity 3D SLAM maps and just-in-time post-processing generation of dense RGB-D 3D maps for enhanced scene representation.

The UAV system integrates two extremely low-cost sensors: a 4K RGB camera and a 3D time-of-flight sensor. The combination enables autonomous human detection, classification, and precise localization within the generated maps. Additionally, our software incorporates autonomous path-planning capabilities with obstacle-avoidance functionality.

Key Features of EpiSci's Solution:

• Computationally Efficient SW for Small, Low-Cost Blue UAS Platform
  • Our software runs on any x86 or ARM platforms, including ModalAI VOXL2 boards featuring 8-core processing
  • Compact, lightweight UAV design ensures easy portability and rapid deployment
  • 360-degree protective airframe enables safe and agile operations in confined spaces
  • Real-Time Operations:
    • HD video streaming from the front-facing camera to the operator
    • Live transmission of high-fidelity 3D mapping data
    • Autonomous human detection and object localization
    • Dynamic path planning with collision avoidance
    • Post-Processing Capabilities
    • Enhanced 3D environmental mapping through AI algorithms
    • Advanced scene reconstruction and analysis
• Communications Infrastructure:
  • Multi-drone mesh networking for extended-range operations
  • Encrypted communication protocols
• Advanced System Features:
  • GPS-denied navigation capabilities
  • Low-light and zero-light operational modes
  • Auto-flip recovery system (Turtle Mode)
  • Blue/Green UAS and Ground Control Station compatibility
  • Cost-effective solution (< $10K BOM for multi-drone package)
  • Two-way audio transmission capability

Figure 1 (below): UAV solution leverages ModalAI VOXL2 board

 

Figure 2. System Performance Comparison (below): Left: Real-time video feed from our UAV captured during the final demonstration. Right: Our post-processed Dense RGB-D SLAM 3D Map for enhanced 3D environmental mapping with detailed terrain representation.

The side-by-side comparison highlights our system's ability to provide both immediate situational awareness through live video streaming and comprehensive environmental modeling through advanced 3D mapping algorithms.

 

Figure 3 (below): Left: an example of the test lane built to execute NIST Stage 2.2. Middle: RGB-D Fusion map streamed in real-time during the mission. Right: Dense RGB-D SLAM after post-processing operations.

 

Figure 4 (below): Example of real-world autonomous human detection and positioning on the map. In the picture,  our system autonomously detects people on the video stream (Left), and a green box is placed on the map after the person's location is calculated within the map (Right).

UAS 3.2: LifeLink solution summary

Team EpiSci used an Inova Avenger Pro drone, equipped with a Microhard pX2 and a 2.4 GHz omni-directional antenna, to provide a hotspot for First Responders. The drone connects to a ground station through a Microhard pDDL900 data link and is controlled through a custom QGC GUI which allows the operator to create a LifeLink mission. This mission allows the user to identify multiple receiver locations and use multiple drones. The SW clusters the receiver locations and assigns drones to visit each cluster. The drones continuously cycle through the receivers within a cluster until the user terminates the mission or a failsafe behavior kicks in. Additionally, the user can specify how long the drone visits each receiver, whether a receiver should be skipped, or to pause at a receiver. There are safety mechanisms integrated into the mission for flight deconfliction and terrain checking. Lastly, their drone has foldable propellers and legs for easy deployment and storage.

Team description

EpiSci is a multidisciplinary innovation-driven company that develops next generation autonomous technologies for defense, aerospace, and commercial applications. EpiSci’s diverse workforce includes advanced AI researchers in the areas of SLAM, navigation, exploration, computer vision, communication, signal processing, and networking, to provide both domain expertise and the ability to deliver solutions for both research-oriented and operation-aware autonomy. With a portfolio of Department of Defense and commercial technology, EpiSci develops and deploys trustable, modular, secure, and resilient autonomous systems to increase operator effectiveness while lowering cost and risk.