The use of Autonomous Mobile Robots (AMRs) in livestock facilities is attracting growing interest for inspection and monitoring tasks, as it can reduce the need for direct human intervention in demanding indoor environments. Their deployment, however, requires navigation systems that can operate reliably in constrained and partially structured spaces. The robot must plan and follow a feasible route while responding safely to unexpected static and dynamic obstacles. This thesis presents a modular autonomous navigation framework developed in Robot Operating System 2 (ROS 2) and implemented on a TurtleBot3-based mobile robot. The framework combines a global planning layer, which generates a map-based reference path, with a local planning layer, which converts that path into motion commands while responding to changes in the surrounding environment. For global planning, the environment is represented as a navigation graph derived from occupancy grids and costmaps. Dijkstra’s algorithm establishes the shortest-path foundations, while A*, Theta*, and D* Lite are implemented and compared to examine the trade-offs between search efficiency, path geometry, and adaptation to map changes. For local planning, Pure Pursuit (PP) provides a geometric path-tracking baseline. Its combination with the Vector Field Histogram (VFH) adds reactive obstacle avoidance, while the Dynamic Window Approach (DWA) and a Model Predictive Control (MPC)-based planner provide alternative approaches to motion generation under the robot’s kinematic constraints. The framework is evaluated in simulated static and dynamic navigation scenarios, including tests with obstacles absent from the map. The results compare the planning methods and show how the interaction between global and local planning affects path following and obstacle avoidance. Together, these findings provide an experimental basis for modular AMR navigation in indoor inspection tasks.

The use of Autonomous Mobile Robots (AMRs) in livestock facilities is attracting growing interest for inspection and monitoring tasks, as it can reduce the need for direct human intervention in demanding indoor environments. Their deployment, however, requires navigation systems that can operate reliably in constrained and partially structured spaces. The robot must plan and follow a feasible route while responding safely to unexpected static and dynamic obstacles. This thesis presents a modular autonomous navigation framework developed in Robot Operating System 2 (ROS 2) and implemented on a TurtleBot3-based mobile robot. The framework combines a global planning layer, which generates a map-based reference path, with a local planning layer, which converts that path into motion commands while responding to changes in the surrounding environment. For global planning, the environment is represented as a navigation graph derived from occupancy grids and costmaps. Dijkstra’s algorithm establishes the shortest-path foundations, while A*, Theta*, and D* Lite are implemented and compared to examine the trade-offs between search efficiency, path geometry, and adaptation to map changes. For local planning, Pure Pursuit (PP) provides a geometric path-tracking baseline. Its combination with the Vector Field Histogram (VFH) adds reactive obstacle avoidance, while the Dynamic Window Approach (DWA) and a Model Predictive Control (MPC)-based planner provide alternative approaches to motion generation under the robot’s kinematic constraints. The framework is evaluated in simulated static and dynamic navigation scenarios, including tests with obstacles absent from the map. The results compare the planning methods and show how the interaction between global and local planning affects path following and obstacle avoidance. Together, these findings provide an experimental basis for modular AMR navigation in indoor inspection tasks.

Development of an Autonomous Navigation Framework for Indoor Inspection Robots in Livestock Facilities

PERINI, ANDREA
2025/2026

Abstract

The use of Autonomous Mobile Robots (AMRs) in livestock facilities is attracting growing interest for inspection and monitoring tasks, as it can reduce the need for direct human intervention in demanding indoor environments. Their deployment, however, requires navigation systems that can operate reliably in constrained and partially structured spaces. The robot must plan and follow a feasible route while responding safely to unexpected static and dynamic obstacles. This thesis presents a modular autonomous navigation framework developed in Robot Operating System 2 (ROS 2) and implemented on a TurtleBot3-based mobile robot. The framework combines a global planning layer, which generates a map-based reference path, with a local planning layer, which converts that path into motion commands while responding to changes in the surrounding environment. For global planning, the environment is represented as a navigation graph derived from occupancy grids and costmaps. Dijkstra’s algorithm establishes the shortest-path foundations, while A*, Theta*, and D* Lite are implemented and compared to examine the trade-offs between search efficiency, path geometry, and adaptation to map changes. For local planning, Pure Pursuit (PP) provides a geometric path-tracking baseline. Its combination with the Vector Field Histogram (VFH) adds reactive obstacle avoidance, while the Dynamic Window Approach (DWA) and a Model Predictive Control (MPC)-based planner provide alternative approaches to motion generation under the robot’s kinematic constraints. The framework is evaluated in simulated static and dynamic navigation scenarios, including tests with obstacles absent from the map. The results compare the planning methods and show how the interaction between global and local planning affects path following and obstacle avoidance. Together, these findings provide an experimental basis for modular AMR navigation in indoor inspection tasks.
2025
Development of an Autonomous Navigation Framework for Indoor Inspection Robots in Livestock Facilities
The use of Autonomous Mobile Robots (AMRs) in livestock facilities is attracting growing interest for inspection and monitoring tasks, as it can reduce the need for direct human intervention in demanding indoor environments. Their deployment, however, requires navigation systems that can operate reliably in constrained and partially structured spaces. The robot must plan and follow a feasible route while responding safely to unexpected static and dynamic obstacles. This thesis presents a modular autonomous navigation framework developed in Robot Operating System 2 (ROS 2) and implemented on a TurtleBot3-based mobile robot. The framework combines a global planning layer, which generates a map-based reference path, with a local planning layer, which converts that path into motion commands while responding to changes in the surrounding environment. For global planning, the environment is represented as a navigation graph derived from occupancy grids and costmaps. Dijkstra’s algorithm establishes the shortest-path foundations, while A*, Theta*, and D* Lite are implemented and compared to examine the trade-offs between search efficiency, path geometry, and adaptation to map changes. For local planning, Pure Pursuit (PP) provides a geometric path-tracking baseline. Its combination with the Vector Field Histogram (VFH) adds reactive obstacle avoidance, while the Dynamic Window Approach (DWA) and a Model Predictive Control (MPC)-based planner provide alternative approaches to motion generation under the robot’s kinematic constraints. The framework is evaluated in simulated static and dynamic navigation scenarios, including tests with obstacles absent from the map. The results compare the planning methods and show how the interaction between global and local planning affects path following and obstacle avoidance. Together, these findings provide an experimental basis for modular AMR navigation in indoor inspection tasks.
AMR
Path Planning
Obstacle Avoidance
Control
Trajectory Tracking
File in questo prodotto:
File Dimensione Formato  
Perini_Andrea.pdf

Accesso riservato

Dimensione 8.92 MB
Formato Adobe PDF
8.92 MB Adobe PDF

The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.12608/116380