Autonomous Driving of Industrial Tractor-Trailers Vehicles

In this project, we would like to build from scratch an autonomous industrial tractor-trailers system to automatically transport stuff in industrial scenarios like warehouses, airports and cargo terminals. I have been contributing to this project during my whole Ph.D. career. I've gained theoretical and hands-on experience in its whole development process: retrofit, electrical and sensor system, software and computing system, modeling, identification, perception, localization, planning and control.

We built the vehicle and the vehicle built me either. I have been trained to be a robotics guy with in-depth understanding of robotic modeling, state estimation (localization & system identification), motion planning, motion control and thier real-life implementations on physical vehicles. My academic research and engineering works mainly contributed to its Localization (Configuration Estimation), Motion Planning, Modeling, System Identification and Motion Control.

Sensor Layout and System Architecture

Modeling and Motion Control

Different from existing works in the literature that only consider the kinematics, we account for the dynamic model.

We install the force sensor to simplify the dynamic model and meanwhile improve the motion control performance.

We design the trajectory tracking controller using back-stepping techniques.

Modeling of the whole tractor-trailers vehicle

Simplified model with force sensors

Hongchao Zhao, Zhe Liu, Zhiqiang Li, Shunbo Zhou, Wen Chen, Chuanzhe Suo and Yunhui Liu, “Modelling and Dynamic Tracking Control of Industrial Vehicles with Tractor-trailer Structure,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst. (IROS), pp. 2905-2910, Macau, China, 2019.

Hongchao Zhao, Shunbo Zhou, Wen Chen, Zhiqiang Miao and Yunhui Liu, "Modeling and Motion Control of Industrial Tractor–Trailers Vehicles Using Force Compensation," in IEEE/ASME Transactions on Mechatronics (T-MECH), vol. 26, no. 2, pp. 645-656, April 2021, doi: 10.1109/TMECH.2021.3057898.

Localization and Motion Planning

Instead of localizing single mobile unit, we precisely localize in real time the lead tractor and all the trailers behind with pose-graph optimization.

The online planned trajectories are obstacle-free, dynamically-feasible and, most importantly, enjoy smaller trajectory tracking errors in real execution.

Instead of being applied in computer simulations or scale-down vehicles like many existing works, our methods are successfully applied in the full-size, real-life industrial tractor-trailers system with satisfactory performance.

Test run towing two full trailers with obstacles in the way.

Test run towing three full trailers with obstacles in the way.

Hongchao Zhao, Wen Chen, Shunbo Zhou, Zhe Liu, Fan Zheng, and Yunhui Liu, “Online Trajectory Planning for an Industrial Tractor Towing Multiple Full Trailers,” in Proc. IEEE Int. Conf. Robot. Autom. (ICRA), pp. 6089-6095, Paris, France, 2020.

Hongchao Zhao, W. Chen, S. Zhou, F. Zheng and Y. -H. Liu, "Localization and Motion Planning of Industrial Tractor–Trailers Vehicles," in IEEE Transactions on Control Systems Technology(T-CST), vol. 31, no. 6, pp. 2928-2936, Nov. 2023, doi: 10.1109/TCST.2023.3275497.

Vehicle System Identification

We have been doing system identification through all the development stages and continuously improving the accuracy.

We would like to identify, with only on-tractor sensors, the effective driving wheel radius, ratio between the steering wheel and the front wheels, position of the Center of Gravity, moment of inertia, tire stiffnesses, etc.

These are important values for improved localization, planning and control.

A RAL journal paper is to disseminate our contribution to this part:

Hongchao Zhao, Wen Chen, Shunbo Zhou and Yunhui Liu, "Parameter Estimation of an Industrial Car-Like Tractor," in IEEE Robotics and Automation Letters (RA-L), vol. 6, no. 3, pp. 4480-4487, July 2021, doi: 10.1109/LRA.2021.3068943.