Researchers in China are finding ways around the limitations of UAV battery life for large-scale monitoring applications (writes Nick Flaherty).

As missions cannot be completed by a single UAV powered by one battery (typically lasting 30-40 minutes), owing to the wide scope of inspections required on a construction project, the scheduling of UAVs of multiple types and their charging demands are key considerations.

Construction environments vary dynamically, so multi-round inspection tasks must be conducted to search for hidden dangers. To prevent collisions between UAVs and facilities, no-fly zones are needed to limit the routes that UAVs can use when travelling between sites, making path-planning challenging.

The classical vehicle routing (VRP) algos used for logistics and delivery services do not work for large monitoring applications, say researchers at Shanghai University in China, who are developing alternatives.

In conventional VRP scenarios, UAVs usually visit each customer point once within a known time window, with no instances of network inaccessibility between customer points. However, construction monitoring needs UAVs to make multiple visits to specific points within periodic time windows with no-fly zones, and consider the charging locations.

A mathematical model and a tailored algo were developed for planning UAV inspection routes to search for hidden dangers on an engineering project.

Consecutive inspection tasks for a given site are separated by a time window, and a mixed-integer linear programming (MILP) model is used to consider the limitations of the batteries and no-fly zones.

This model determines the routing and scheduling of multiple different UAVs with varying periodic inspection and charging cycles and no-fly-zone constraints. A tailored algo, based on a variable neighbourhood search (VNS), was used to solve the model at large scale.

This is being used for inspection at the Shiziyang Bridge project in Guangdong province, which is being built with a world-record main span of 2180 m.