Delft University of Technology
Faculty Mechanical, Maritime and Materials Engineering
Transport Technology



T.F. ter Hoeven A study of the NOMAD model and its possible application on AGV systems
Literature survey, Report 2004.TL.6904, Transport Engineering and Logistics.


AGV systems have been around for almost fifty years. Although many improvements have been made, AGV systems still use fixed driving routes. When AGVs cross each others driving path, this often causes delays in the system. It is therefore very interesting to determine whether an AGV system with more freedom of path choice performs better than the traditional systems. Such a system will have to be controlled by a model that finds the optimal paths to destinations whilst avoiding collisions between AGVs.
Serge Hoogendoorn, a researcher at the faculty of Civil Engineering at the Technical University of Delft, has developed a microscopic pedestrian model called NOMAD. This model has been used to aid in the design of public spaces such as airport terminals and train stations. The model considers pedestrians to be economists, in a way that the interests of a pedestrian can be seen as costs or utility, which can be optimized. The question is whether this model can be adapted for use as a control model of an AGV system.

NOMAD consists of three levels; the strategic, tactical and operational levels.
At the strategic level, several activity sets are created. These activity sets act as inputs for the tactical level, where a simultaneous optimization decision is made for the activity schedule and time, destination area and route choice. The optimal driving route is determined by optimizing the so-called running costs. These consist of the factors travel time, proximity to obstacles, energy use due to walking speeds, density of the pedestrian field and stimulating effects of the environment. Every running cost factor has a weighting factor assigned, which enables the user to set up the model to plan routes in the way that is demanded for a certain situation. The output of the tactical level is an optimal velocity path describing the route to a certain destination. At the operational level, a similar approach is used to determine the behaviour of pedestrians while following the path that is determined at the tactical level. Again, an optimum is sought for the running cost factors, which consist of straying from the optimal velocity path, proximity to other pedestrians or obstacles, and energy use due to acceleration or deceleration. The behaviour of the pedestrians can be changed by assigning different values to the individual weighting factors of the running costs.

Application of the NOMAD-model on the AGV systems of container terminals is an interesting prospect. By changing the cost and weighting factors of both levels of the model, the behaviour of the pedestrians of Hoogendoorn's model can be adapted to suit the character of AGVs. The question remains whether an AGV system with this type of control outperforms the more traditional type of control.
Simulating an AGV container system will increase our insight in the use of the NOMAD model; because of the bi-level structure of the model, it is possible to simulate the tactical and operational levels separately. A logical first step would be to set up a simple simulation with fixed driving paths to study how the behavioural level of the model copes with AGV encounters. With this experience, a more elaborate simulation could be set up to create a simulation for the tactical level.


Reports on Transport Engineering and Logistics (in Dutch)
Modified: 2005.03.11; logistics@3mE.tudelft.nl , TU Delft / 3mE / TT / LT.