|
|
Ymro Hoogendoorn (Erasmus University) - Sensor placement in a single-period waste collection problem Supervisor: Rommert Dekker and Albert Wagelmans Recorded full presentation Abstract A recent development in waste collection is the placement of sensors to reduce uncertainty. However, as placing sensors require both installation and upkeep costs, we want to make an informed decision to place sensors. In addition, sensors are not perfectly accurate and we have to choose between cheaper sensors that give imprecise readings, or more expensive sensors that give precise readings. Given the readings of the placed sensors, the single-period waste collection problem reduces to a vehicle routing problem with stochastic demands (VRPSD). We allow for correlated demand. The VRPSD can be solved exactly by means of an integer L-shaped method. However, to evaluate different sensor placements, the expected costs of a single sensor placement needs to be calculated, which is equivalent to computing the expected cost of many different optimal VRPSD solutions. In this research, we derive theoretical properties of this placement problem, and devise different methods to approximate or upper bound the expected costs of a sensor placement. These expected costs are then used to infer strategies for sensor placements using instances found in waste collection practice. |
