TRASH: Tandem Rover and Aerial Scrap Harvester

TL;DR

This paper introduces an autonomous multi-robot system combining aerial and ground robots with CNN-based detection for efficient highway litter monitoring and collection, demonstrating promising results in simulations and real-world trials.

Contribution

The novel integration of aerial scanning with ground-based litter detection and a greedy pickup strategy advances highway cleanup automation.

Findings

Effective litter detection and mapping in real-world trials

Successful autonomous navigation to litter locations

Potential for improved highway cleanliness management

Abstract

Addressing the challenge of roadside litter in the United States, which has traditionally relied on costly and ineffective manual cleanup methods, this paper presents an autonomous multi-robot system for highway litter monitoring and collection. Our solution integrates an aerial vehicle to scan and gather data across highway stretches with a terrestrial robot equipped with a Convolutional Neural Network (CNN) for litter detection and mapping. Upon detecting litter, the ground robot navigates to each pinpointed location, re-assesses the vicinity, and employs a "greedy pickup" approach to address potential mapping inaccuracies or litter misplacements. Through simulation studies and real-world robotic trials, this work highlights the potential of our proposed system for highway cleanliness and management in the context of Robotics, Automation, and Artificial Intelligence