Collision Avoidance in Cluttered Environments: A Low-cost Mechatronic Approach for Autonomous Robots.

Abstract

The development of a collision avoidance system for autonomous robots is essential to ensure the safety of both the robot and its operating environment. This abstract examines the key elements and technologies involved in designing an effective system tailored for autonomous ground vehicles (AGVs). As autonomous technology progresses, the demand for increased computational resources grows. However, the emergence of affordable single-board computers (SBCs) with robust processing capabilities has opened new possibilities. This thesis is driven by the goal of creating an economical navigation and obstacle avoidance solution.

The study explores the use of various sensors, imaging devices, and sophisticated algorithms to identify potential obstacles and provide real-time alerts to the SBC. It also covers the integration of safety features such as autonomous emergency braking, workspace boundary alerts, and adaptive speed control to reduce collision risks. The sensor system aims to equip an AGV with sufficient environmental data, enabling safe and efficient operation in dynamic settings. To achieve this, the project established core system requirements at the outset:

1. The system must deliver comprehensive data over a wide area, detecting all obstacles within its range to allow timely AGV responses and prevent collisions.

2. It should dynamically update the environmental map as conditions change.

3. It must include a mechanism to continuously monitor the AGV’s position within its operational space.

In this thesis, “low-cost” is defined as a budget suitable for laboratory use with institutional support, targeting a total cost of approximately $500. The design prioritizes ease of integration with existing robotic platforms. Leveraging community support and available software libraries, the system is built using the ROS robotic framework, specifically the ROS2 Dashing distribution, alongside the Dynamic Window Approach (DWA) algorithm. The chosen SBC for this project is the Raspberry Pi 4 Model B (RBPi4).