Collaborative Research: NSF-DST: RI: Small: Algorithms from wetware:A neural model of spatial navigation in very large and cluttered space

When navigating in complex environments, fixed landmarks and moving obstacles are crucial features that influence efficient and robust path planning, optimal route finding, and minimization of navigational errors. Autonomous vehicles are severely limited by their inability to reliably anchor their navigation to landmarks and predict and avoid the movement of others. The research team proposes to develop and refine a computational model of spatial navigation and spatial representation using neural data obtained wirelessly from animals navigating in the two largest electrophysiology-compatible rodent mazes in the world, which are known as “megaspaces.” These studies explore the influence of stationary landmarks and moving objects as rats optimize their routes: a classic paradigm (the Traveling Salesperson Problem) in Computer Science. In addition to their technological impact in robotics and autonomous vehicles, these investigations can be extended to human mental health dysfunctions that are often accompanied by deficits in spatial processing such as in early onset Alzheimer’s disease, attentional deficit hyperactivity disorders, schizophrenia, or depression. This investigation is novel and unique in trying to understand how the interactions between the hippocampus and entorhinal cortex, two main components of the brain’s ‘GPS’ system, facilitate navigation, learning, and complex decision making in very large spaces. The research involves using experimental data to constrain a detailed biophysical neural model and testing experimentally its predictions about the properties of neural representations of megaspaces in challenging navigational tasks. The model will provide a new tool for the detailed study of the use of fixed landmark and moving obstacles in very large environments for efficient navigation. The work will contribute to robotics and computational neuroscience along two different axes: (1) using data-constrained modeling to propose concrete mechanisms explaining the nature of the interactions between self-motion and landmark-based navigational information and (2) using neural representations of large space to achieve efficient solutions or approximations for generally hard spatial navigational problems that could have significant impact in many disciplines. A companion project is being funded by the Department of Science and Technology, India. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.