Autonomous Truck with a Trailer

Project Overview 2023

The Autonomous Truck With Trailer project at Linköping University, part of the TSRT10 course, aims to develop an educational and research platform in autonomous vehicles. The 2023 project focused on enhancing the LEGO trucks LQ and MPC controller, to handle dynamic obstacles. This included implementing advanced motion planning and control systems to navigate around moving objects like pedestrians and vehicles, using a predictor based on the Extended Kalman Filter for obstacle anticipation.

Controller

In this project, two types of controllers, an MPC and an LQ, manage steering and speed. The MPC utilizes a fast Cvxpy solver to avoid redundancy in data processing. While the LQ controller is a legacy system from past projects, it has been significantly refined and integrated into the new framework for enhanced user accessibility.

Dynamic Obstacle prediction

Dynamic obstacles in the system are modeled to represent pedestrians and ground vehicles, two common variables in real-world scenarios. The system employs a Predictor that utilizes an Extended Kalman Filter to forecast the future positions of these moving obstacles. Additionally, there is flexibility for manual adjustment by users to reposition the obstacles as needed.

Motion Planning

The motion planner's role is to chart a viable path to the target. It's equipped to replot the route to navigate around obstacles. If obstacles are encountered, it will pause briefly, allowing for their potential clearance before rerouting. Obstacles that are distant along the path are initially disregarded until they fall within a specific lookahead threshold.

Results

System Integration and Performance

The autonomous vehicle system showcased good performance in mission execution. The core components, including the Model Predictive Controller (MPC), Obstacle Simulator, and Predictor, have been adeptly integrated within the Robot Operating System (ROS) framework.

Motion Planning Proficiency

Separately, the Motion Planner has been crafted within a C++ milieu and integrated in the ROS framework this year. The Motion Planner has been tested in a variety of scenarios in the Qualisys projector room, and has shown good performance in avoiding static and dynamic obstacles.

Foxglove and Qualisys

The Autonomous Truck project has made significant progress in integration with Foxglove and using the Qualisys projector room at Linköping University. This integration allows for realistic testing environments, enhancing the project's effectiveness in real-world scenarios.