Rowdy Robots -- Autonomous Wet Gap Crossing Agents

Abstract

Terrain has historically and currently presented issues in transporting and obtaining resources, ammunition, and rations through their proper supply channels. For example: rivers, streams, lakes, and entire oceans continue to be obstacles faced by militaries. The challenge of determining how to successfully traverse through these geographical barricades in an efficient and safe manner also persists. Due to the fact that terrain is used strategically in crucial events, it is imperative to be able to cross varying terrains in situations without delay as seconds could saveÑ or costÑthousands of lives. Therefore, having a system that could aid in crossing any water-based geographical feature that is full-proof, rugged, and adaptive to the situation is crucial for an advanced military. In order to address this problem, we propose a network of remote-controlled and AI assisted robots that will autonomously and simultaneously construct a bridge, provide cover fire, and help navigate vehicles across the constructed bridge safely and swiftly. This advancement would free up the soldiers that would otherwise be manning the boats to push segments of the ribbon bridge into place and the personnel required to attach these parts together. This allows the soldiers more freedom to focus on other necessary tasks and priorities, such as ensuring the safety of others or neutralizing threats. While these robots will be equipped with onboard computers capable of autonomously processing situations with multiple sensors and making decisions in fractions of seconds, users will also have the option of controlling these robots manually. The autonomous action of the robots is the result of trained neural networks. The AI will be trained off-site on better equipped computational devices that will drill and optimize algorithms that can be loaded onto the onboard computers of the robots. This will ensure safe data-protection of vital AI information housed in protected data centers and ease the load of the onboard computers, giving them more headroom for sensors and faster processing times. Real-time, onboard training of AI will continue to be a struggle with current hardware limitations. As time progresses AI will be able learn from new situations, growing and deepening its neural networks. Until onboard training is more feasible, new information gathered while a system is deployed can be sent back to high performance data centers, analyzed, and the resulting updates will be sent back to the robot.

Document Details

Document Type
DoD Grant Award
Publication Date
Mar 24, 2020
Source ID
W911NF2010041

Entities

People

  • Patrick Benavidez

Organizations

  • Army Contracting Command
  • Office of the Secretary of Defense
  • University of Texas at San Antonio

Tags

Readers

  • Materials Science
  • Robotics and Automation.
  • Strategic Security Studies

Technology Areas

  • AI & ML
  • AI & ML - Autonomous Systems
  • AI & ML - DoD AI Strategy
  • AI & ML - Neural Networks
  • Autonomy