Ohio State Doctorate Student Works on Autonomous Vehicle Emergency Maneuvering Systems

An Ohio State doctorate student is developing a system of controllers that would allow autonomous vehicles to react to split-second emergency situations faster and more accurately than a human driver could.

Vivek Bithar is a doctorate student at The Ohio State University in the Department of Mechanical and Aerospace Engineering working under Dr. Shawn Midlam-Mohler in the autonomous vehicle space. Bithar’s research combines two objectives, path planning and tracking when focusing on the kind of situations where a vehicle must perform emergency maneuvers.

Path planning and control are an important part of autonomous vehicle software systems. The autonomous vehicle has to decide on a path and then using its ‘actuators’ which are functions such as steering and braking to adjust itself and follow that path.

“My dissertation is trying to combine these two objectives. Whenever you see a certain kind of emergency maneuver required, the system which I’ve designed is called into action; if you’re driving in a straight line and then a pedestrian walks in front of you or the vehicle ahead of you slows down suddenly, the system is activated and then using the reference path it will optimize it’s trajectory and re-plan it around that obstacle,” Bithar said.

Bithar’s software will improve upon current path planning and tracking systems which often lead to inaccuracies and errors when it comes to real world maneuvering.

 “The whole objective is to avoid obstacles in a robust fashion, which means taking into account the inaccuracies in the modeling and the friction, avoiding the obstacle and bringing the vehicle back on its original path; and on a road where there are no obstacles around, have it follow its path, which was its basic objective,” Bithar said.

The autonomous vehicle controllers need to be really fast since they are working on emergency maneuvering situations. The controllers, combining the objectives of the vehicle and being robust to the uncertainty of the environment, precompute much of their calculations in an offline database, and then optimize the trajectory using the number from this database online to meet real-time requirements. Often, these controllers are not adaptive to changes in the models and are computationally inefficient, resulting in errors and performance limitations.

“Currently in emergency maneuvers the path planning of the vehicles layer uses non sophisticated models which leads to many errors when trying to follow these paths, eventually there will be a collision. They can never guarantee a collision free path. When the controller tries to track that path there are some errors,” Bithar said.

Bithar’s research tackles the inaccuracies and errors that come with precomputing calculations in an offline database, his research focuses on a completely online method called ‘fast online tube based MPC’.

 “The other way to do it is without offline computation at all. Everything is done online within the time that you have and that’s actually a major challenge- it’s called fast online tube based MPC,” Bithar said. “Fast online tube based MPC has never been tried in this situation. There is a process section called linearization, which is the approximation of non-linear models. When you approximate them, the offline computations become invalid for such an implementation because the model changes due to linearization at every execution step of the controller, which is the advantage of doing all of these calculations online in real-time using fast online MPC.” Bithar said.

This research is a major contribution to the autonomous vehicle space. Current autonomous vehicles on the road are considered “level 2” driver assisted systems, Bithar is developing technology considered “level 4 or 5” which exceed the abilities of a human driver.

Written by Muhammed Al Refai, CAR Marketing and Communications Intern