Reactive collision avoidance for Articulated Frame Steering vehicle: Articulated frame steering occupancy grid mapping control barrier function velocity safety filter
Kärki, Topi Reino Johannes (2025)
2025
Konetekniikan DI-ohjelma - Master's Programme in Mechanical Engineering
Tekniikan ja luonnontieteiden tiedekunta - Faculty of Engineering and Natural Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2025-08-29
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202508288555
https://urn.fi/URN:NBN:fi:tuni-202508288555
Tiivistelmä
Mobile working machines are a set of diverse industrial machines with a wide variety of possible working environments and work tasks. One shared task between all mobile working machines is the traversal of the environment, and in many cases this environment can be complex and subject to very frequent change. This poses a major safety risk and a challenge for automation.
This thesis extends previous work, where the authors propose a method of safe traversal for initially unknown environments using sensor-based environmental modeling and mathematical formulation of machine state safety. Occupancy grid mapping (OGM) is used to define the model of the environment using perception sensors. A control barrier function (CBF) is used to guarantee the safety of the system by defining a function that restricts the allowed states of the machine based on environmental factors.
This thesis proposes and implements changes to the original method with the aim of generalizing it for use as a velocity command safety filter for an articulated frame steering vehicle for the purpose of collision avoidance in unknown environment traversal. The environmental modeling is expanded with Lagrangian interpolation to mitigate the impact of the discrete nature of OGM with the mathematical constraints of the CBF formulation.
The goal of the thesis work is to test the proposed safety filter implementation using an ideal kinematic simulator and a real autonomous articulated frame steering mobile working machine. The test results from the simulator are analyzed to identify the performance of the safety filter on a theoretical level and to tune the hyperparameters. Real machine tests are used to assess the practical performance of the naive kinematical implementation in real-life use.
The modifications made for the implementation of the method on an articulated frame steering vehicle are successful, and the Lagrangian interpolation positively impacts the behavior of the safety filter. The method still has instances of breaking theoretical CBF constraints, while still successfully performing collision avoidance in all real-world tests.
This thesis extends previous work, where the authors propose a method of safe traversal for initially unknown environments using sensor-based environmental modeling and mathematical formulation of machine state safety. Occupancy grid mapping (OGM) is used to define the model of the environment using perception sensors. A control barrier function (CBF) is used to guarantee the safety of the system by defining a function that restricts the allowed states of the machine based on environmental factors.
This thesis proposes and implements changes to the original method with the aim of generalizing it for use as a velocity command safety filter for an articulated frame steering vehicle for the purpose of collision avoidance in unknown environment traversal. The environmental modeling is expanded with Lagrangian interpolation to mitigate the impact of the discrete nature of OGM with the mathematical constraints of the CBF formulation.
The goal of the thesis work is to test the proposed safety filter implementation using an ideal kinematic simulator and a real autonomous articulated frame steering mobile working machine. The test results from the simulator are analyzed to identify the performance of the safety filter on a theoretical level and to tune the hyperparameters. Real machine tests are used to assess the practical performance of the naive kinematical implementation in real-life use.
The modifications made for the implementation of the method on an articulated frame steering vehicle are successful, and the Lagrangian interpolation positively impacts the behavior of the safety filter. The method still has instances of breaking theoretical CBF constraints, while still successfully performing collision avoidance in all real-world tests.