College - Author 1

College of Engineering

Department - Author 1

Electrical Engineering Department

Degree Name - Author 1

BS in Electrical Engineering

College - Author 2

College of Engineering

Department - Author 2

Electrical Engineering Department

Degree - Author 2

BS in Electrical Engineering

College - Author 3

College of Engineering

Department - Author 3

Electrical Engineering Department

Degree - Author 3

BS in Electrical Engineering

Date

7-2026

Primary Advisor

Noel Ellis, College of Engineering, Electrical Engineering Department

Additional Advisors

Jenna Kloosterman, College of Engineering, Electrical Engineering Department Chuck Bland, College of Engineering, Electrical Engineering Department

Abstract/Summary

AAC Vision designed and constructed a vision-based autonomous tracking rover capable of exploring an indoor environment, avoiding obstacles, generating a two-dimensional LiDAR map, detecting a standard orange basketball, and approaching the target without manual steering. The rover uses a Raspberry Pi 4, an Intel RealSense D435 RGB-D camera, an RPLIDAR C1, four JGB37-520 Hall-encoder motors, dual H-bridge motor drivers, a 12 V battery, and a custom aluminum chassis measuring approximately 8 in by 12 in and weighing 6.7 lb. The final mission emphasized reliable target discovery rather than travel between predetermined points. The software integrates Python, OpenCV, Ultralytics YOLO, RealSense depth processing, wheel-encoder odometry, BreezySLAM occupancy mapping, ROS 2 in- terfaces, and reactive obstacle avoidance. Target detection combines neural-network classification with orange-color, shape, depth, position, and temporal-consistency checks. The approach controller aligns the target in the camera image, advances in bounded steps, rechecks depth after each movement, and stops near a commanded stand-off distance of 0.80 m. Development required many software revisions because changes that improved one behavior often exposed another: permissive thresholds produced false positives, continuous motion caused target loss, aggressive recovery produced inefficient turns, and conservative navigation increased mission time. Testing demonstrated successful sensor integration, saved basketball detections, autonomous target approach in favorable trials, partial room mapping, and reliable obstacle avoidance in open layouts. The most persistent limitations were corner traps, incomplete room coverage, wheel-slip effects on odometry, and Raspberry Pi undervoltage under some power configurations. The project demonstrates the feasibility of combining semantic vision and geometric mapping on a low-cost embedded rover while identifying clear improvements for power distribution, localization, global planning, and formal validation.

Share

COinS