Search results

The purpose of this paper is to achieve intelligent superstore site selection. Yonghui Superstores partnered with Cardinal Operations to incorporate a tremendous amount of site-related information (e.g. points of interest, population density and features, distribution of competitors, transportation, commercial ecosystem, existing own-store network) into its store site optimization.

This paper showcases the integration of regression, optimization and machine learning approaches in site selection, which has proven practical and effective.

The result was the development of the “Yonghui Intelligent Site Selection System” that includes three modules: business district scoring, intelligent site engine and precision sales forecasting. The application of this system helps to significantly reduce the labor force required to visit and investigate all potential sites, circumvent the pitfalls associated with possibly biased experience or intuition-based decision making and achieve the same population coverage as competitors while needing only half the number of stores as its competitors.

To our knowledge, this project is among the first to integrate regression, optimization and machine learning approaches in site selection. There is innovation in optimization techniques.

The purpose of this paper is to enable autonomous door-opening with unknown geometrical constraints. Door-opening is a common action needed for mobile manipulators to perform rescue operation. However, it remains difficult for them to handle it in real rescue environments. The major difficulties of rescue manipulation involve contradiction between unknown geometrical constraints and limited sensors because of extreme physical constraints.

A method for estimating the unknown door geometrical parameters using coordinate transformation of the end-effector with visual teleoperation assists is proposed. A trajectory planning algorithm is developed using geometrical parameters from the proposed method.

The relevant experiments are also conducted using a manipulator suited to extreme physical constraints to open a real door with a locked latch and unknown geometrical parameters, which demonstrates the validity and efficiency of the proposed approach.

This is a novel method for estimating the unknown door geometrical parameters with coordinate transformation of the end-effector through visual teleoperation assists.

Geometric errors are common in metallic bent tubular parts. Thus, tubes should be inspected and fixed before welding with the joints first. After welding, the relative position of the joints is also necessary to be inspected to judge whether the tube can be assembled reliably. Therefore, the inspection plays an important role in the tube’s assembly. The purpose of this paper is to propose a multi-vision-based system designed to inspect the tube and the relative position of the joints.

For the tube inspection, the small cylinders are taken as the primitives to reconstruct the tube using the multi- vision-based system. Then, any geometric error in the tube can be inspected by comparing the reconstructed models and designed ones. For joints’ inspection, authors designed an adapter with marked points, by which the system can calculate the relative position of the joints.

The reconstruction idea can recognise the line and arc segments of a tube automatically and resolve the textureless deficiency of the tube’s surface. The joints’ inspection method is simple in operation, and any kinds of joints can be inspected by designing the structure of the adapters accordingly.

By experimental verification, the inspection precision of the proposed system was 0.17 mm; the inspection time was within 2 min. Thus, the system developed can inspect a tube effectively and automatically. Moreover, authors can determine how the springback of the arcs behaves, allowing in-process springback prediction and compensation, which can reduce geometric errors in the tubes given the present bending machine accuracy.