An Online Optimization Algorithm for Path Planning of Unmanned Rollers for Compaction of Dams

The unmanned roller, as an important technical means to reduce manual labor intensity, can improve the operation efficiency and quality. Due to the change of road surface conditions, it is difficult to drive in a straight-line accurately. As a result, some area that needs to be rolled missed. So it is necessary to select appropriate repeated rolling width (referred to as overlapping distance) for adjacent strips according to real-time rolling conditions. Therefore, in this paper, a self-learning online optimization method is proposed, to optimize the overlapping distance in the cloud computing center. To this end, a multi-objective cost function is proposed, considering impacts of overlapping distance on missed rolling percentage, over-rolling percentage and operation efficiency, which turns the problem of overlapping distance online adjustment into a convex optimization problem. In the cloud side, with the minimum cost function as the goal, the best overlapping distance is searched in real-time by using the mountain-climb searching algorithm, and then sent to the realtime control system of unmanned roller. Finally, the proposed algorithm is validated in experiments. Results show that compared with the fixed overlapping distance algorithm, the over-rolling ratio can be reduced by 4.05% and 5.06%, the fuel consumption can be reduced by 4.10% and 1.83% in the flat transition zone and the large-size rockfill zone respectively.