Robotic Following using Deep Learning

JMP proposes to develop a system for robotic following using deep learning in a way that has not been done before by using 1) a novel hybrid CNN-RNN-encoder-decoder-based end-to-end deep learning approach to perform real-time scene analysis from passive video camera images, including lead vehicle identification, and to make time-series aware driving decisions, 2) a three-phase deep learning training protocol combining unsupervised, supervised, and reinforcement learning phases that supports lifelong incremental learning in a way that does not over-fit to specific training sets and is robust to environments and conditions that are different from the training, and 3) a proof-of-concept testbed for collecting large-scale real-world data sets and performing real-world tests using deep learning mobile robots.