A Novel Seal-Whisker-Inspired Cyber-Physical Hydrodynamic Sensing System for Complex Flow Detection, Perception and Tracking (ONR White Paper Tracking Number: 23-000005714)
Abstract
Approved for Public Release.This work presents an interdisciplinary basic research endeavor that aims at (1) extracting the fundamental principles of biomechanics and fluid mechanics that underlie the remarkable flow sensing abilities of seal whiskers, including object detection, identification, and tracking, and (2) building upon the gained insights, designing a cutting-edge cyber-physical hydrodynamic sensing system to enable autonomous underwater detection, identification, and tracking with high sensitivity, accuracy, and intelligence. By integrating this advanced sensing system into autonomous underwater and amphibious vehicles, we envision a substantial enhancement in their manipulation capabilities, characterized by intelligent object perception and tracking, ultimately expanding the operational envelope of Navy underwater vehicles. Three objectives are proposed toward the goal.Objective 1. Develop a seal whisker-inspired sensory system with advanced capabilities for sensing subtle and complex ambient flow features. Focused features include high sensitivity, short response time, strong repeatability, long durability, and multi-directional sensing. Particularly, we propose to develop novel 3D-printed microelectromechanical systems (MEMS) with nanoparticle silver pizoresistor through state-of-the-art stereolithography and aerosol printing technology to achieve high sensitivity of each individual whisker. We will also explore bio-inspired whisker array architecture to capture comprehensive flow information. Through a combination of experiments in water tanks and computational simulations, we will design, optimize, and verify the sensory system. Objective 2. Develop an embodied AI system for underwater vehicles with bio-inspired perception and tracking. This involves creating an interpretable machine learning solution to identify and capture key characteristics of vortex structures in the wake, enabling intelligent object perception (type, shape, size, location, etc.) Leveraging this perception, a value-based deep reinforcement learning algorithm will be designed for prey-like navigation, utilizing vortical wakes to locate target objects. In particular, the algorithm will guide the underwater vehicle to actively collect vortical information and intelligently follow the hydrodynamic wake trails based on the sensory feedback of ambient flow information. These advanced perception and tracking capabilities pave the way for complete automation of underwater vehicles. Objective 3. Uncover the flow sensing mechanisms of seal whiskers by investigating the following hypotheses: (a) wake vortex structures provide unique signatures of moving objects in water, (b) individual whiskers sense local flow velocity, and (c) collective signals from whisker arrays enable the perception of complex vortex structure information (size, speed, orientation and moving direction). Seals use these vortex structures to track objects upstream. Multidisciplinary approach encompassing computational, theoretical, experimental and machine learning methods will be used. We will establish comprehensive relationships between local flow velocity and individual whisker root mechanical signals, as well as the connections between complex vortex structure information and whisker array collective signals. Particularly, we will leverage video transformer-based machine learning algorithms to examine if arrays signals can successfully learn crucial vortex structure information and identify key signal patterns associated with them. The improved understanding will inform the design of objective 1 and 2.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Nov 21, 2023
- Source ID
- N000142412023
Entities
People
- Xudong Zheng
Organizations
- Office of Naval Research
- Rochester Institute of Technology
- United States Navy