Project Grant 2223822

This four-year, $1.96 million project grant from the National Science Foundation's Office of Emerging Frontiers and Multidisciplinary Activities, under the Engineering program (CFDA 47.041), will develop improved brain-machine interface algorithms to restore movement and speech abilities. The University of California, San Diego will apply advances in understanding biological motor control to design decoding strategies enabling robust prosthetic control across contexts. Existing feedforward neural network decoders demonstrating initial success in controlled environments will be augmented with theoretically and empirically motivated features of the biological motor system. Brain-machine interface algorithm design will be accelerated using validated modular recurrent neural networks simulating brain dynamics in motor areas. Algorithms will be evaluated against existing nonhuman primate and songbird neurophysiology data, and validated through new in vivo experiments collecting wider behavioral manipulation data to test learning and generalization. The research aims to co-develop neuroscience theory and algorithms across species and contexts to produce superior brain-machine interface decoding algorithms and uncover principles for developing machine learning approaches interacting efficiently and flexibly with intelligent controllers.