Proceedings of the 2017 Winter Simulation Conference
W. K. V. Chan, A. D’Ambrogio, G. Zacharewicz, N. Mustafee, G. Wainer, and E. Page, eds.
AN AGENT-BASED SIMULATION MODEL FOR AUTONOMOUS TRAILER DOCKING
Berry Gerrits
Martijn Mes
Peter Schuur
Department of Industrial Engineering and Business Information Systems
University of Twente
P.O. Box 217
7500 AE Enschede, THE NETHERLANDS
ABSTRACT
This paper presents a simulation model of a generic automated planning and control system for the pick-up
and docking of semi-trailers by means of autonomous Yard Tractors (YTs) in a collision- and conflict free
environment. To support the planning and control of the YTs, we propose a Multi-Agent System (MAS).
We illustrate our approach using a case study at a Dutch logistics service provider. To evaluate the proposed
system, we design an agent-based simulation model, which is set up in a similar way as the MAS. We
conclude with the verification and validation of the simulation model.
1 INTRODUCTION
The past few years have witnessed an increased interest within the logistics sector for automated driving.
For logistics service providers it yields an opportunity to decrease waiting time and maneuvering time at
DCs and thus to overcome problems with the driver’s DOT (Department of Transportation) rules. For a
truck unloading at a distribution center, there are various ways of implementing automated driving. We
distinguish between two options once the truck arrives: (i) the truck driver is assisted using an automated
driving system or (ii) the truck and semi-trailer are decoupled and an Automated Guided Vehicle (AGV),
a so called Yard Tractor (YT), takes the semi-trailer to the dock. This paper focuses on the latter. Our aim
is to design (i) a stable and robust planning and control system and (ii) an agent-based simulation model
to evaluate this system.
For the planning and control of transportation resources, in this case the YT, we could resort to
operations research (OR) based global optimization methods. However, these methods may be less suitable
for real-time decision making in stochastic and dynamic environments, since they typically (i) require a lot
of information in advance, (ii) are sensitive to information updates, (iii) are not able to respond in a timely
manner, and (iv) are not flexible enough to deal with changing environments and situations with multiple
autonomous actors (Mes, van der Heijden, and van Harten 2007). An alternative is to use a distributed
planning approach in the form of a Multi-Agent System (MAS). A MAS consists of several independent
and autonomous control units (agents) that pursue their own interests and interact with other agents in
the environment. MASs are believed to be particularly suitable for decentralized systems in real-time and
dynamic environments (Mes, van der Heijden, and van Hillegersberg 2008).
One of the key design challenges of MASs, is the configuration of the agents such that their self-
interested behavior yields a near-optimal solution for the network as a whole. The network is defined as a
closed transportation network consisting of a fixed number of pick-up and drop-off (P/D) locations. YTs
transport the goods (e.g., semi-trailers) between these locations using a certain track layout. The network
is closed, so no YT can enter or leave the system, even when idling. Similarly to Ebben (2001), this
1324 978-1-5386-3428-8/17/$31.00 ©2017 IEEE