MORE | Spring 2021
Online Prediction for Vision-based Active Pursuit using a Domain Agnostic Offline Motion Model
The use of a Long Short-Term Memory Network-based domain-agnostic predictive pursuit agent is proposed as an alternative to conventional methods such as Kalman Filtering. The empirical results from the pursuit-evasion game establish the superiority of the proposed framework as attested to by lower capture times. This active pursuit framework enhances the ability of autonomous vehicles to navigate rapidly changing environments and situations. Future work involves enhancing the perception capabilities of the pursuing agent and the deployment and coordination of multiple pursuers.
Hometown: Pusad, Maharashtra, India
Graduation date: Spring 2021