Abstract
This project presents the development of TALOS, an autonomous robot designed for use in server rooms. TALOS was built to handle a surprisingly human but critical task: plugging and unplugging ethernet cables during server downtimes. By automating this process, TALOS reduces manual labor, lowers the risk of human error, and increases the efficiency of server room operations. Equipped with a sophisticated gripper, autonomous navigation, and computer vision–based port detection, TALOS demonstrates the potential of robotics in data center reliability and automation.
Introduction
Server rooms are the backbone of modern organizations. But when a server goes down, troubleshooting often involves manual intervention — plugging and unplugging ethernet cables, tracing network connections, and rerouting lines. These tasks are tedious, error-prone, and slow, leading to costly downtime.
The TALOS project set out to build a robot that could autonomously perform cable management in server rooms, freeing up human operators and reducing downtime risks.
Motivation
Automate repetitive and error-prone tasks in server rooms.
Improve reliability and uptime in data centers.
Explore how robotics can bring automation to traditionally overlooked environments.
System Overview
Proposed Solution
TALOS was designed to autonomously navigate a server room.
It uses computer vision to detect LAN ports and plan the optimal path.
A custom gripper enables safe plugging/unplugging of ethernet cables.
Admins can interact with the robot through a front-end system, sending retrieval and target addresses for cable connections.
Gripper & Manipulation
Designed a sophisticated cable gripper that can securely hold ethernet connectors.
Tested on a robotic arm under simulated motion to ensure reliability.
Capable of repeated plug/unplug operations without slip.
SCARA Arm Design
Developed a custom SCARA arm, 3D printed and assembled as part of TALOS.
Integrated with the gripper to reach ports in server racks.
Computer Vision
QR Code Detection: Used Python-based libraries for QR recognition, transmitting results into ROS for robot decision-making.
LAN Port Detection:
Built a custom dataset of LAN ports with different notch orientations.
Classified into four categories: Up, Down, Left, Right.
Trained a CNN in Keras, deployed as an
.h5model for real-time port detection with confidence scores.
Software & Architecture
Integrated ROS for communication and control.
Developed a front-end interface for admins to interact with TALOS.
Implemented navigation and path planning for autonomous movement.
Prototype Development
Designed and fabricated the SCARA arm with gripper integration.
Mounted gripper on a robotic arm for testing stability.
Verified QR code scanning and LAN port CNN classification.
Demonstrated cable plug/unplug cycles in lab tests.
Results
The gripper successfully gripped and released ethernet cables without slip.
CNN-based LAN port detection classified port orientation reliably.
QR code–based identification enabled accurate retrieval and routing of cable destinations.
Front-end system allowed seamless interaction between admins and robot.
Conclusion
The TALOS project successfully demonstrated the feasibility of using an autonomous robot for cable management in server rooms. By combining robotics, computer vision, and custom gripper design, the team built a working prototype that can navigate, detect ports, and handle ethernet cables safely.
TALOS shows how automation can extend beyond warehouses and factories into data center operations — reducing downtime, minimizing human error, and improving overall reliability. With further development, TALOS could evolve into a deployable robot for modern server farms.
