I-Corps: Neuromorphic Target Tracking and Control for Insect-Scale Aerial Vehicles

Cornell University, Ithaca NY

Investigators

Abstract

The broader impact/commercial potential of this I-Corps project is to enable a broader range of end users to utilize the capabilities of autonomous robotics and to make advanced autonomous systems more broadly accessible and reliable. Robotics such as micro aerial vehicles (MAVs), self-driving cars, and other ground-based service robots are playing increasingly important roles in the lives of many industries as advances in autonomy enable them to be used safely and reliably in diverse situations. Accurately tracking targets and navigating while avoiding obstacles are important prerequisites to fully autonomous operation for robots. Neuromorphic cameras can more accurately detect motion than traditional cameras while consuming far less power. This project will explore the commercial applications of algorithms which interpret the data from neuromorphic cameras to enable autonomous systems to accurately track motion and navigate in unknown environments. By enabling more accurate sensing with reduced power consumption, these algorithms will increase the safety and reliability of autonomous systems. The proposed techniques will enable autonomous systems to react safely and robustly in real time to unexpected environmental changes without immediate operator intervention. This I-Corps project will explore the commercialization of neuromorphic sensing and control algorithms that enable accurate environmental sensing from moving robotic platforms. Autonomous navigation requires processing data from exteroceptive sensors for the purposes of obstacle avoidance and target tracking. These tasks must be accomplished in real time with minimal latency to maximize the capabilities and reliability of the autonomous robot. Neuromorphic cameras sense the environment with sub-millisecond latency and, unlike traditional cameras, provide information only about changes in the scene. The algorithms which will be explored by this project efficiently process the data from neuromorphic sensors to detect the presence of moving targets and stationary obstacles to enable efficient autonomous control for high-speed aerial and ground-based robots in uncertain and rapidly changing environments. These algorithms include neuromorphic control techniques which enable autonomous control in the presence of both environmental uncertainties such as disturbances and uncertain variations in the physical parameters of the robot. The proposed techniques have been validated in high-fidelity simulations using benchmark datasets and have been shown to be capable of rapid adaptation to unexpected changes while maintaining control of the autonomous system. 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.