In-Motion Weight Sensor Array for Dynamic Weighing of Nonsingulated Objects

This paper considers a problem of in situ estimation of the weight values of individual items that are moving as a bulk on a conveyor in a nonoriented and nonsingulated way. In logistics and material handling processes, it is a challenge to measure the weight of each package moving on a conveyor without halting the conveyor operation. This is not only because measuring weight requires physical contact between the target item and the measuring instrument, but also because each item must be separated (or singulated) from others for it to be weighed. Such a singulation task causes a significant delay in the transport process, especially when the conveyor is heavily loaded with items. Because of this issue, despite the importance of weight information for correct pricing and safe loading (for both air and ground vehicles), the weight has largely been ignored in many logistical sites such as sorting facilities or distribution centers. To address these issues, this paper proposes a new instrumentation device called the in-motion weight sensor (IMWS) array that consists of a grid of mini-checkweighers operating separately from each other as independent weight measurement cells. Each IMWS unit is equipped with a mini-conveyor mounted on a load cell, and provides the composite load data for a group of items being carried within the IMWS system. A depth camera mounted above the IMWS surface identifies both the location and the orientation of each package and provides the real-time measurement of its partial areas that intersect with the surface of load cells in contact with each package. The load cell measurements and the vision data are then fused to construct the measurement equation that is linear in the unknown weight values of all the packages being transported within the IMWS system. To extract individual weight values from these composite load data, we implement the recursive least square (RLS) estimation to the IMWS array. The feasibility of the proposed idea has been verified with experimental tests using the full-scale prototype IMWS array system. The results show that IMWS array is capable of estimating the unknown weight with the mean error around ±3% error and the standard deviation of 9 ~13% error for cases where 3 ~ 5 randomly oriented packages are moving in a nonsingulated way.