Sketching for Efficient and Robust Top-Down Autonomous Navigation

The award supports a research project that introduces a novel approach that robots can use to navigate (individually or as part of a team) in unknown environments in a way that is closer to how humans approach the same task. The main project novelty is represented by the high-level reasoning used by the robots (which uses rooms and doors instead of simple points and planes as in previous approaches) for both keeping track of the environment during exploration, and for deciding how to maneuver.

The project’s impacts are twofold: 1) the methods will make robotic systems more robust, as they will not rely on specific details of an environment but on a more “general” understanding of it, and 2) the techniques and code developed will be the basis for a new modular software teaching platform to be used in the Robotics and Autonomous Systems Teaching and Innovation Center (RASTIC) to educate the next generations of workers in robotics and engineering.

On a technical level, the project is organized along three components: 1) Compact, interpretable, high-level representations of the environment based on structured elements that are: directly extracted from measurements using machine learning model, compared using Signed Distance Functions, and optimized over time using piecewise linear optimization. 2) Local and global navigation strategies that synthesize nonlinear controllers via linear programming, and stitch them together using Q functions that take into account uncertainty. 3) Coordination approaches for robot teams that reduce sensing requirements to a barebone set of measurements (contact sensors, single-beam rangefinders, compass), and avoid computation during navigation by reusing previous experience. Overall, the goal is to provide new approaches for mapping and control synthesis that do not require detailed bottom-up representations of the environment and that use a fraction of the computational requirements of previous techniques.

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.