TECHNOLOGY AREA(S): Sensors
OBJECTIVE: The goal of this topic would be to improve off-road autonomous mobility in military environments as mentioned above using relatively low-cost or COTS sensors while combining them with novel memory techniques.
DESCRIPTION: Recent advancements in sensors and processing have significantly improved the capabilities of autonomous ground vehicles, particularly in the commercial market. The military environment poses several unique problems to Robotic Autonomous Systems (RAS) including incomplete or insufficient map data, dynamically changing terrains, and Global Positioning System (GPS)/communications denied environments that could increase the time to complete a mission or cause mission failure. Novel processing algorithms that include machine learning and artificial intelligence could increase the speed of ground RAS and decrease the likelihood of mission failure. They may require "training" of the RAS through either supervised or unsupervised techniques on a representative area.
PHASE I: The vendor will conduct necessary tradeoff studies/analyses of conventional versus proposed techniques of robotic maneuver to prove feasibility and capability of the proposed approach.
PHASE II: Design and build a prototype ground or air robotic navigation system with increased capability for demonstration in several varied environments (benign open terrain, wooded site, urban, indoor, and in GPS challenged environments).
PHASE III: The vendor will commercialize the system. Military application of this topic is directly applicable to Army robotics efforts via the Assured PNT program, subprogram Mounted PNT. Commercial applications of this technology would be also directly applicable to First Responders (fire fighters, police, security, and other emergency units), hobbyists, and for telecommunications/infrastructure inspection.
1: Pieter Abbeel, Adam Coates, Timothy Hunter, Morgan Quigley and Andrew Ng, Helicopters teach themselves to do aerial maneuvers", http://news.stanford.edu/news/2008/september10/helicopter-091008.html Proceedings of the 20th annual conference on Computer graphics and interactive techniques, p.73-80, August 2008
2: Sergey Levine, Peter Pastor, Alex Krizhevsky, Deirdre Quillen; Learning Hand-Eye Coordination for Robotic Grasping with Deep Learning and Large-Scale Data Collection, arXiv:1603.02199, http://arxiv.org/abs/1603.0219, Mar 2016.
KEYWORDS: Autonomy, Artificial Intelligence, Machine Learning, Positioning, Navigation, PNT