This work is part of the SmartSite research project (http://smartsite- project.de) that aims at the optimization of the entire logistical chain of black-top assembly. It is envisioned to integrate our proposed software agent as an assistance system into a broader construction site logistics management and operations system once the simulated experiments have tested successfully. Although SmartSite focuses on the entire logis- tics chain, this article addresses only the control of compactors.

The remainder of the article is structured as follows. In Section 2 we provide a brief overview of the current state-of-the-art in the construc- tion of hot-mix asphalt, assistance systems for compactor operators, and multi-agent systems. While Section 3 describes our artefact in detail, Section 4 introduces the evaluation scenarios. Results of the simulation and field tests are analyzed in Section 5. Section 6 concludes the paper.

2. State-of-the-art

2.1. The black-top assembly  phase

A road construction site during the black-top assembly phase is a highly complex logistical system influenced by a large amount of envi- ronmental, technical, social, and managerial factors. To get the right amount of asphalt on time in the right quality (i.e., the right tempera- ture) to the paver and “on the road” at minimum costs is a quite challenging task. The logistics chain incorporates asphalt batch plants, transportation facilities, pavers and compactors (see Fig. 2).

Each step is susceptible to numerous disruptions and the paving crew on site as the focal point of the whole system often lacks informa- tion about disruptions at upstream logistical steps. Research and prac- tice concerning this problem are based on proprietary, isolated, and in parts heterogeneous data structures [6]. Furthermore, current solutions are  focused  on  performance  monitoring  and  on  documentation

subsequent to the construction phase rather than on supporting machine operators during construction and on an operational level. These shortcomings have a notably effect especially on the compaction of asphalt.

As the last step in the logistics chain, compaction is of particular im- portance for the quality of the final road. An inappropriate number of passes leads either to over- or under-compaction of the asphalt. For the human compactor operators it is often difficult and impossible to estimate the right number of passes remaining. Even if all areas are passed an appropriate number of times, too low core temperatures of the asphalt during the pass can lead to an under-compaction or damage to the asphalt surface. If compactors pass too hot areas, liquid compo- nents of the bitumen are pressed out and there is an additional risk of subsidence. The right number of passes within the right temperature range (influenced by environmental and material-related conditions) is the crucial factor for quality. Operators are not adequately supported by appropriate methods and systems and the desired density after compaction is often not reached (see Fig. 3).

2.2. Related work in the field of assistance systems for compactor operators

The literature provides several approaches for equipment operator assistance systems and automated compactors. These approaches main- ly rely on pre-planned path calculations and lack flexible decision and reaction on changing conditions [5,7,8]. Other related approaches, e.g. automated surface crack detection and autonomous sealing of asphalt pavement, focus on road maintenance rather than road construction [9,10]. Real-time sensory inputs are often not considered [7,11] and information about environmental conditions are not incorporated in the path calculations [12–14]. Table 1 shows some of the recent approaches in assistance systems for compactor operators within the literature and reports about evaluations results as well as differences to our approach.