Research

Energy-Aware Multi-Agent Task Allocation, Scheduling, and Planning

The goal of this project is to come up with a planning algorithm for a team of vehicles that maintains consistent energy in a team: multiple ground vehicles will support multiple aerial vehicles with recharging capabilities. Both type of vehicles must visit as many target locations as possible and respond to dynamic events.

S*: A Framework Generalizing Bidirectional Search to Solve Multi-Goal Path Finding

In "Multi-goal path finding", we wish to find the shortest path visiting each goal once, beginning from a source and ending at a destination. The problem is NP-Hard and there exists 2-approximation algorithms. We apply heuristic bidirectional search to improve the efficiency of such algorithms. The result is S*

Chour, K., Rathinam, S., and Ravi, R. "S*: A Heuristic Information-Based Approximation Framework for Multi-Goal Path Finding," In Proceedings of the International Conference on Automated Planning and Scheduling, 2021, May, Vol. 31, pp. 85-93.

Aerial-Ground Vehicle Path Coordination Algorithms

The problem is defined as follows: A ground vehicle (GV) needs to travel along a narrow corridor (for example a highway), but the surrounding area is unknown. There are however aerial vehicles available for the GV, which can be used to survey the surrounding area. Several high priority targets (in white) are also known. The objective is for the AVs to strike a balance between low and high-priority coverage during survey. All vehicles will converge at some point further along the corridor, where information is transferred to the GV. The process repeats until the GV has reached its destination.

Chour, K. and Rathinam, S.,"Aerial-ground vehicle path coordination algorithms for monitoring applications," Artificial Intelligence and Machine Learning for Multi-Domain Operations Applications II.Vol. 11413. International Society for Optics and Photonics, Mar. 2020

Improving Autonomous Vehicle Cybersecurity via Dynamic Watermarking

The goal of this project was to implement "Dynamic watermarking", a technique used to improve cybersecurity among networked systems, on a full-size autonomous vehicle. By injecting a secret signal into the control output (from drive-by-wire system), we should ideally be able to recover it from the sensor. This allows us to detect spoofing attacks on the sensor and then issue a defensive response such as stopping the car.

Path Following Algorithms for Autonomous Vehicles

I have implemented several path following algorithms based on pid, pure pursuit, and sliding mode control. The vehicles I have worked on include the Polaris XP 900, Jeep Grand Cherokee, and Lincoln MKZ. As part of this, I became familiar with ROS and programming with Python/C++.

Liu, M.,Chour, K., Sivakumar, R., Darbha, S. "Lateral Control of an Autonomous and Connected Vehicle with Limited Preview Information," IEEE Intelligent Transportation Systems, Oct. 2020

Hassen, S.,Chour, K., Weaver, A. and Gopalswamy, S., 2020. "Nonlinear Control of a Ground Vehicle using Data-Driven Dynamic Models," (No. 2020-01-0171) SAE Technical Paper.

Infrastructure Enabled Autonomy (IEA)

The goal of this project was to distribute sensing and control onto nearby infrastructure. This would allow autonomous vehicles to avoid carrying expensive sensors like LIDAR. Several multi-sensor smart packages (MSSP) were installed on light posts at RELLIS, a testing campus about 25 minutes from Texas A&M University.