Autonomous Vehicles for Intelligent Mobility

Here at the UNT College of Engineering, our research in connected ground vehicles spans across a wide array of disciplines and applications. With 8 faculty members leading graduate and undergraduate students, our research intersects the disciplines of electrical, mechanical, engineering technology, materials science, and both computer science and computer engineering.

Connected Autonomous Vehicles (CAV)

Accurate pedestrian and cyclist detection using 2D and 3D data fusion
Finding blurry vehicles using Generative Adversarial Network
Designing efficient deep learning models for autonomous vehicles
Cooperative perception with data sharing between cars
Edge computing for traffic monitoring

Edge computing

Diagram showing edge computing structure

Edge servers are much closer to cars
Low latency for data movement, high responsiveness
More computing capability than cars
Local traffic management
Traffic monitoring & update, accident analysis, intersection safety

Finding blurry vehicles

Using enhanced super resolution Generative Adversarial Network (SRGAN)

blurry vehicle detection

Cooperative perception

LiDAR based data fusion for connected and autonomous vehiclesLiDAR based data fusion for connected and autonomous vehicles

Goals of data fusion

Enhances the accuracy of object detection for autonomous vehicles
Provides real-time information for traffic management and traffic control
Provides anonymity of private information within the sensing range
Increases system’s reliability in case of sensor failures

Object detection on fused frames

Vehicle #1
Vehicle #2
Fused result

Occluded road sign detection

LiDAR-Assisted Detection

Transfer learning based solution

Experiment results

Low-power hardware for real-time edge ML

A fully programmable architecture based on the open RISCV ISA, leveraging:
- Deep Quantization
- Feature sparsity
- Vectorized Execution
- Memory-side scatter-gather
Feedback-driven Framework for ML ISA

Vehicle location privacy protection

Vehicle traffic-aware location privacy threat models
Location obfuscation mechanisms to protect against traffic-aware inference attacks
Time-efficient obfuscation generation algorithms

Block diagram of Obfuscation framework and threat model

Person and object detection

Images with color squares on detected objects.

Smart microgrids for autonomous vehicle power

Block diagram of low voltage smart microgrids

Bottom-up power electronics microgrid solution
Smart energy utilization technology
Scalable, cost-effective and environmental-friendly design
Hybrid circuit breaker solution for long life time protection

Mesh network for swarm robotics

The swarm robotics can achieve more challenging goals than what any single robot can achieve by working collaboratively.
Individual robots may have simple design with limited computing, power, and functional capabilities.
The swarm robotics network is distributed, decentralized, self-organizing, self-healing, and scalable.

BLE mesh network implementation with ESP32
for localization and tracking applications
TI-RSLK MAX robotic cars as swarm robotics platform

Autonomous Ground Vehicles (AGV) optimization and precision manufacturing

Frame and parts design — Design and Optimization of Automotive Structures

Finished prototype parts — Rapid Prototyping and Proof of Concept Testing

Three equipment in Digital Manufacturing Lab — Additive Manufacturing Systems

Drone network and communications

Aerial Communication Infrastructure for Smart Emergency Response

First drone-carried WiFi communication system
Long-range (up to 5 miles) and broadband (54Mbps) drone-to-drone wireless link
Received various awards and media coverage
Community-centric project with collaborators: NCTCOG, Austin, Denton, Tarrant

Networked Airborne Computing Platform

Smart platform of modular design and full functionality
Flexibility and extensibility for new development
Friendly application development capability

Drone with comp, ctrl and comm three units

3D Communication System

Robust communication between drones for coordination
Fast data collection and information fusion with drone networks and on-ground sensors
Broadband networks including air-to-air and air-to-ground to enable real-time information exchange
Seamless integration of drone networks with Internet of Things (IoT)