Title: Advancing Stochastic Trajectory Optimization to Achieve Human-Level Capabilities

Date: Monday, February 16, 2026

Time: 9:30AM - 10:30AM ET

Location:  CODA 1215 Midtown

Virtual Link:  https://gatech.zoom.us/j/99310170679
Meeting ID: 993 1017 0679

Passcode: 9088308

Bogdan Vlahov

Ph.D. Robotics Student

School of Mechanical Engineering

Georgia Institute of Technology

Committee:

Dr. Evangelos Theodorou (Advisor)

School of Aerospace Engineering

Georgia Institute of Technology

Dr. Glen Chou

School of Aerospace Engineering

Georgia Institute of Technology

·  Dr. Frank Dellaert

School of Interactive Computing

Georgia Institute of Technology

Dr. Lu Gan

School of Aerospace Engineering

Georgia Institute of Technology

Dr. Animesh Garg
Department of Interactive Computing

·  Georgia Institute of Technology

· 

Abstract:

Over the past several years, optimization has become a crucial component to our daily lives. With a focus on stochastic trajectory optimization,  a variety of methods to improve the current state-of-the-art and achieve human-like capabilities in the field of autonomous off-road driving are presented. These methods include a new sampling distribution for the Model Predictive Path Integral (MPPI) algorithm, a software library for MPPI that provides control task flexibility and improved performance, and a new hierarchical planning architecture that takes into account dynamical uncertainty, perceptual uncertainty, and approximation biases. These methods have been tested in real-time hardware applications with great success. Finally, new adaptation schemes to better balance the differences between the different planners as well as machine learning techniques to improve the tuning of the cost function and controller for autonomous driving will be proposed.