The heavy truck with its trailers is backed around the corner and then manoeuvred into the right position at the loading bay. There is, however, no driver behind the steering wheel – the operation is fully automated.
“Our field experiments with a truck with a dolly and a trailer show that the system performs well, with very high repeatability. The current focus is not on extreme manoeuvres, where I judge that experienced drivers perform better. However, there is no fundamental limitation in the results to prevent this from changing in the future, although it will place higher demands on sensors and actuators”, says Oskar Ljungqvist.
Motion Planner
In the work described in his licentiate thesis, Oskar Ljungqvist has developed what is known as a motion planner, which makes it possible to instruct the vehicle to move from one point to another, and to end up at a desired position. It may involve manoeuvres in both forward and backward motion, and it is, of course, important to keep track of the long trailer.“The planned manoeuvres should preferably be as intuitive as possible. If it is possible to solve the task by driving forwards, the system will do so”, he says.
The motion planner uses a map of the surrounding environment in which various known obstacles are included. New obstacles are detected by sensors on the vehicle. Based on this information, the planner calculates the best motion plan to reach the desired goal location. Sensors located on the truck enable the system to keep track of the truck, the trailer and the joint angles between them.
“The motion planning algorithm uses a number of precomputed base manoeuvres, which are combined in different ways during online planning to solve the task”, says Oskar Ljungqvist.
A LIDAR sensor on the truck emits and detects laser light reflected from the body of the trailer. Using these observations, distances and angles between the truck, dolly and trailer can then be estimated.
Few control signals
“One of the difficulties here is that there are few control signals compared to the complexity of vehicle. The joint angles between the truck and the trailer are particularly important – it is of great importance that the vehicle does not end up in a jack-knife configuration”, says Oskar Ljungqvist.If an unexpected object, such as another vehicle suddenly appears, the autonomous vehicle stops and the system calculates a new optimal path to the goal.
Some parts of the system were developed in a previous project, iQMatic, financed by Vinnova. Oskar Ljungqvist’s interest for the subject was stimulated when he worked on his master’s thesis project within iQMatic, where he was inspired to continue to research education. After his master’s in applied physics and electrical engineering at LiU, he continued initially to work within the iQMatic project, and then continued to another project financed by FFI: Stability and controller structures for self-driving vehicles. Both projects have been conducted in collaboration with Scania.
“Starting with the work from iQMatic, Oskar Ljungqvist has extended it with functions to enable the truck to plan and execute relatively complex manoeuvres with trailers”, says Daniel Axehill, associate professor at the Division of Automatic Control, and Oskar Ljungqvist’s main supervisor.
Oskar Ljungqvist shows theoretically in his licentiate thesis that the vehicle can be guaranteed to follow the path it has calculated to be optimal, and he has developed new methods for how this can be verified. The framework is built on methods developed in the fields of robotics, mathematical optimization and automatic control.
From theory and Lego to test track
“It has been a long project, and we have learnt a lot during the process. We started with computer simulations, progressed to laboratory testing with a truck built from Lego, and finally performed field experiments on a full-scale test vehicle on Scania’s test track. The system has undergone continuous improvement during the project, and new theoriesHe has only good things to say about the collaboration with Scania:
“They have been very helpful; it has been a learning experience for me, and I have spent many hours at their test track, not least together with Henrik Pettersson.”
The movie available on a link below was recorded at the test track.
Now that the licentiate thesis has been presented and defended, Oskar Ljungqvist is planning the next steps. One direction for future work is to investigate how learning can be incorporated to improve the overall system performance. The Swedish Strategic Vehicle Research and Innovation (FFI) programme finances his doctoral research, and he is an affiliated PhD student within WASP, the Wallenberg AI Autonomous Systems and Software Program.
Footnote: LIDAR is an acronym for “light detection and ranging” – an optical sensor that measure distances, angles, and other parameters.
The licentiate thesis: On motion planning and control for truck and trailer systems, Oskar Ljungqvist, Department of Electrical Engineering, LiU. Supervisor: Docent Daniel Axehill.
Translation George Farrants
A truck with a dolly and a trailer on Scania’s test track
