http://dx.doi.org/10.1016/j.autcon.2016.03.012 0926-5805/© 2016 Elsevier B.V. All rights reserved.

22 R. Kuenzel et al. / Automation in Construction 71 (2016) 21–33

Fig. 1. (a) State-of-the art in asphalt road paving operation might result in (b) poor quality of road condition during its lifecycle.

Errors in judgments of human operators during compaction are in- evitable, since they are not supported by adequate assistance systems. These must work in real-time to gather, process, and communicate sen- sor information about all relevant factors instantly [5].

Researchers and practitioners have developed several technical and non-technical approaches to this problem and introduced them into the current field of practice. The predominant method is to hand over pre- planned and fixed rolling patterns to the operators to aim at operational efficiency and optimal asphalt density after compaction. The main deficit of this solution is its rigidness and inflexibility regarding any environmental and process-related changes.

Furthermore, operators have to judge material-related characteris- tics like actual asphalt temperature and its compressibility on their own. Up until recently, it has been difficult for human operators to determine the remaining number of passes — and sometimes even to remember how often their machines have already compacted a certain area.

Original equipment manufacturers (OEM) as well as special machinery equipment vendors provide a variety of, mostly proprietary, technical solutions to support machine operators in their work, e.g., by mounting a Global Positioning Systems (GPS)-based counter to visual- ize the number of passes. The main focus of market-available solutions, however, is to provide data about temperature and compaction for documentation purposes subsequent to the construction process rather than to support the operators during compaction.

Moreover, current systems are mostly isolated solutions with a spe- cial purpose. They are not integrated with other environmental, material-related and process-related information. These systems are partially suitable for use in applications subjected to operators' assistance. Hence, there is a need for developing a real-time integrated control system to support human machine drivers at an operational level during the compaction phase of a road construction project.

A solution to the problem is to sense, exchange, store, and analyze data about physical and process-related conditions for every machine in use. This is possible with already market-available solutions. We use such solutions to develop an intelligent assistance system for compactor operators based on standard service-based interfaces and a set of pre-planned rolling patterns. It is capable of adjusting itself to changing conditions — no matter if these changes are caused by the environment or by any steps in any of the upstream processes.

The developed system is capable of controlling a compactor. As such it is also capable of assisting human operators where to steer the ma- chine next according to its current position. Recent and most relevant work history as well as compliance to projected work plans, trajectories of other compactors, actual work progress of the paver itself, and envi- ronmental and material-related parameters are also included in the de- cision making.

This article adopts techniques from the field of distributed artificial intelligence – namely multi-agent systems – to contribute a method that automatically generates driving instructions for human compactor

R. Kuenzel et al. / Automation in Construction 71 (2016) 21–33 23

Fig. 2. Logistics chain of a road construction site during black-top assembly.

operators. The method is capable of reacting to variable and changing field-realistic conditions. We conducted a simulated experiment to evaluate the method and to show the feasibility of our artefact.