Real-Time Localization Framework for Autonomous Basketball Robots
Naren Medarametla, Sreejon Mondal
TL;DR
This work tackles real-time self-localization for autonomous basketball robots under Robocon 2025 by proposing a lightweight, monocular, floor-line–based localization framework. It combines a simple preprocessing pipeline that highlights court floor cues with a compact feedforward network to predict the robot’s $(x,y)$ position, achieving an average error of approximately $0.06\text{ m}$. The study compares against CNN backbones like MobileNetV2 and EfficientNet-B0, analyzes explainability via Integrated Gradients, and discusses failed approaches that relied on depth sensing or coarse zone-based estimates. The results indicate that a monocular, floor-image–driven approach can provide accurate, real-time localization suitable for edge deployment, with clear avenues for integrating additional sensors and optimizing for embedded hardware.
Abstract
Localization is a fundamental capability for autonomous robots, enabling them to operate effectively in dynamic environments. In Robocon 2025, accurate and reliable localization is crucial for improving shooting precision, avoiding collisions with other robots, and navigating the competition field efficiently. In this paper, we propose a hybrid localization algorithm that integrates classical techniques with learning based methods that rely solely on visual data from the court's floor to achieve self-localization on the basketball field.