Model Predictive Control based Motion Planning for Autonomous Mobile Robots using Euclidean Signed Distance Maps
Master Defence
Rashmi Dabir
Bosch Research
Tuesday, June 18, 2024, 11:30
Room 01-012, Georges-Köhler-Allee 102, Freiburg 79110, Germany
This thesis focuses on motion planning and collision avoidance for Autonomous Mobile Robots. The collision avoidance constraints are formulated using Euclidean Signed Distance Transform (ESDT), also known as Signed Distance Map. The proposed differentiable constraint formulation can be incorporated into Model Predictive Control and enables the robot to safely navigate through cluttered environments. Using ESDT, we identify the points on the robot that are closest to the obstacles and could result in a collision. These points are then incorporated into the collision avoidance constraints.
This algorithm is implemented by creating a controller plugin package for the ROS2 navigation stack. The goal of this thesis is to assess the performance of the proposed controller through simulations in the Gazebo environment, and experiments conducted on the physical robot. Furthermore, it also evaluates the controller against the standard Model Predictive Path Integral (MPPI) controller in nav2 based on certain performance indices.