IGVC 2026 · UTA · System Online

Navigating the
Unknown

A lane-following, obstacle-aware outdoor robot built for the Intelligent Ground Vehicle Competition — powered by Jetson Orin Nano, TM4C microcontrollers, LiDAR, and YOLO-based object detection.

Explore Systemarrow_downward View Documentationarrow_forward

Compute

Jetson Orin Nano

MCU

TM4C123

Vision

Camera + YOLO

Ranging

2D LiDAR

Safety

Wireless E-Stop

Scroll to descendarrow_downward

01 · Overview

A multi-modal autonomous ground vehicle

Instead of relying on a single sensor, the system fuses vision, ranging, and learned detection to stay stable when any one source gets noisy or unreliable.

The IGVC project focuses on building a real-time autonomous ground vehicle capable of operating in outdoor competition environments. The robot must follow lanes, avoid obstacles, detect stop signs and pedestrians, and respond safely when sensor data is incomplete or noisy.

// Lane info

Cameras

// Obstacles

LiDAR

// Traffic signs

YOLO

// Motor control

TM4C

02 · Capabilities

Core Capabilities

Five subsystems working together to keep the vehicle aware, stable, and safe in unpredictable outdoor environments.

alt_route

Lane Detection

The camera system detects white lane markers using color masks, region-of-interest filtering, contour detection, and line tracking. The output is a steering value that tells the robot whether to turn left, turn right, or continue forward.

Color maskROI filterContour detectionLine tracking

radar

LiDAR Avoidance

Scans the environment in front-left, front-center, and front-right zones. The robot steers away from the closest obstacle.

visibility

YOLO Detection

Identifies stop signs and pedestrians. A stop or pedestrian-yield behavior triggers when detection is consistent and close.

hub

Sensor Fusion

Combines camera-based lane following with LiDAR-based obstacle avoidance. The system remains stable when lane markings are unclear or obstacles appear in the path.

emergency

Safety System

Wireless emergency stop, communication- and sensor-failure handling, and visible emergency lights for state signaling.

03 · Architecture

System Architecture

A layered architecture that separates perception, decision-making, and actuation so each subsystem can run independently without stalling the rest of the robot.

sensorsLAYER_01

Input Layer

Gathers real-time information from the environment and vehicle.

Cameras detect lane markings
LiDAR measures obstacle distances
GPS supports waypoint navigation
Encoders help track movement
E-Stop switch for immediate shutdown

memoryLAYER_02

Processing Layer

Interprets sensor data and decides how the robot should move.

Jetson Orin Nano for vision, LiDAR, and detection
TM4C handles real-time motor control
Shared memory + multiprocessing per subsystem
Control logic fuses lane tracking and avoidance

boltLAYER_03

Output Layer

Executes the robot's movement and safety behavior.

Motor drivers control vehicle motion
Emergency lights communicate robot state
Fail-safe behavior during unsafe conditions

schema

Image Placeholder · System Architecture Diagram

/public/images/architecture-diagram.png

04 · Integration Protocol

High-Level System Flow

Six stages execute every control loop — from sensing the world to guarding against failure.

STEP_01

Sense

Sensors collect real-time data from the environment — cameras, LiDAR, GPS, encoders, and the emergency stop line.

sensors

visibility

STEP_02

Perceive

Camera processing detects lane markings while LiDAR divides the environment into front-left, front-center, and front-right obstacle zones.

STEP_03

Classify

YOLO identifies stop signs and pedestrians. Detections must be consistent and close enough before triggering a behavior.

psychology

hub

STEP_04

Decide

The controller fuses sensor outputs into a single steering and throttle decision, using hysteresis to prevent jitter near thresholds.

STEP_05

Actuate

Motor commands are sent to the TM4C over UART. A serial lock ensures only one final command reaches the motor driver.

bolt

emergency_home

STEP_06

Guard

Safety monitors watch for communication, sensor, and behavior failures and stop the robot before things go wrong.

05 · Crew Manifest

Meet the Team

Six students bringing together embedded systems, computer vision, robotics, controls, and safety engineering to build a competition-ready autonomous ground vehicle.

person

OPR_01

Member 1

Team Lead · Perception

VisionArchitecture

person

OPR_02

Member 2

Embedded Systems · TM4C

TM4CUARTMotor Control

person

OPR_03

Member 3

Control & Navigation

GPSHysteresisFusion

person

OPR_04

Member 4

Safety · Wireless E-Stop

E-StopFail-safe

person

OPR_05

Member 5

Mechanical · Chassis

ChassisMounting

person

OPR_06

Member 6

Power · Integration

BatteryRailsWiring

swipeDrag or use arrows to scroll

06 · Vehicle Gallery

Active Instances

Chassis, electronics bay, sensor stack, wiring, wheels, motor drivers, and outdoor testing environment.

image

VEH_01

Full Robot View

Complete vehicle, front view

image

VEH_02

Side Profile

Chassis and wheel geometry

image

VEH_03

Electronics Bay

Jetson Orin Nano + TM4C stack

image

VEH_04

Sensor Stack

Camera + LiDAR mounting

image

VEH_05

Wheels & Chassis

Drive train + encoders

image

VEH_06

Outdoor Testing

Field test on lane markings

image

VEH_07

Lane Detection

Camera output + overlay

image

VEH_08

LiDAR Scan

Zone-based obstacle readings

01 / 08

07 · Verification

Testing Plan

The robot was tested across progressively complex scenarios, with each scenario layering new requirements on top of the last.

SC_01

Lane Following

The robot maintains alignment using vision-based lane tracking.

Test Goals

checkDetect lane markings
checkStay centered between lanes
checkCorrect steering errors
checkMaintain stable forward movement

SC_02

Obstacle Avoidance

The robot uses LiDAR to detect nearby obstacles and respond safely.

Test Goals

checkDetect obstacles in front-left, center, and front-right zones
checkAvoid obstacles using steering correction
checkStop when the path is unsafe

SC_03

Hybrid Navigation

The robot transitions between camera and LiDAR control depending on sensor confidence.

Test Goals

checkUse camera when lane visibility is strong
checkUse LiDAR as fallback when lane data is weak
checkRecover from partial or complete lane loss
checkAvoid rapid switching between commands

tune

Steering Stability

±2° jitter

bolt

Obstacle Response

< 120 ms

autorenew

Lane Recovery

Continuous

speed

Runtime FPS

Consistent

verified

Detection Reliability

Stop + Ped.

08 · Hard Problems

Engineering Challenges

Four problems that shaped the final system architecture.

sync_problem

Serial Command Conflicts

Multiple scripts could send commands to the TM4C at the same time. Command management and serial locking ensure only one final command reaches the motor controller.

show_chart

Steering Jitter

Values near the threshold caused rapid left/right/forward switching. Hysteresis smooths motion and prevents oscillation near the decision boundary.

blur_on

Sensor Reliability

Camera performance depends on lighting, shadows, lane color, and calibration. LiDAR acts as a fallback when vision data becomes unreliable.

account_tree

Real-Time Integration

Camera, LiDAR, YOLO, and motor control must all run continuously. Multiprocessing lets each subsystem operate independently without freezing the robot.

09 · Demo Day