Allocation of resources in complex transportation systems has properties of resource allocation systems (RAS) with non-sequential processes with flexible routing and use of resources of multiple types at once. Deadlock states in such a system can be avoided by using the Banker's algorithm. Presentation will start by showing simulation of railway nodes in Villon, a tool for detailed discrete simulation models of logistic nodes - as the initial situation for the deadlock problem. In the main part, a sample coloured Petri net (CPN) model of the analyzed RAS and the Banker's algorithm and its necessary modifications for solving the problem will be explained. The model has been implemented in the CPN Tools, software tools for modelling and analysis of CPN, and the algorithm in the CPN ML, adjacent functional language for CPN models in the CPN Tools. The Villon simulation tool uses discrete simulation of the agent-based architecture (ABA). For implementation of agent's behaviour and communication between agents in models, a specific subclass of Petri nets called ABA-graph has been developed. A CPN version of the ABA-graph called ABA-CPN will be presented in the second part of the presentation.