PIBOT: Design and development of an indoor room autonomous mobile robot for HI-VIS safety vest person-following with IR-based obstacle avoidance

Yong, Ivan Chi-Kin (2026) PIBOT: Design and development of an indoor room autonomous mobile robot for HI-VIS safety vest person-following with IR-based obstacle avoidance. Final Year Project, UTAR.

PDF Download (1290Kb)

Abstract

Automation and AI-powered technologies that offer real-time detection and response are slowly replacing traditional systems and workers for better efficiency. The industrial autonomous bots are costly, difficult to operate and can be dangerous while there is no sensor to stop them when obstacles are blocking them. This paper details the design, implementation, and performance evaluation of PiBot, an autonomous mobile robot engineered to follow individuals wearing high-visibility safety vests while navigating around obstacles. It is developed on a Raspberry Pi 3B+ platform, the robot utilizes a modified foldable trolley chassis powered by dual 24 V 120 W DC motors and a 20 V 4 Ah lithium battery pack. Next, the core of the system is a dual stage vision pipeline designed for efficiency on embedded hardware. First is using a MobileNet Single Shot Detector (SSD) initially identifies human targets, while the HSV/LAB colour based tracker maintains a follow rate of 28 frames per second. To ensure operational safety, PiBot integrates two infrared (IR) sensors governed by a high priority interrupt system and a 5 sample debounce algorithm, ensuring the robot prioritizes obstacle avoidance over all other commands. Furthermore, beyond autonomous operation, the project features a custom built, low latency WebSocket control interface. This web-based system achieves an average command round trip time of 14 ms and provides a real time MJPEG video stream at 20 FPS. In conclusion, the findings demonstrate high-performance autonomous navigation, and reliable remote monitoring can be achieved using cost effective, off-theshelf components and optimized software architectures.

Actions (login required)