Delivery Scheduling in City

Modified a concurrent scheduler approach to consider different priorities of tasks, and a practical counter to rotate use of resources for continuous performance analytics to minimize resources. 

Concurrent scheduler approach makes more sense compared to the usual Operations Research approach to scheduling/routing, as this approach is more dynamic, takes conservative traffic prediction into account but is responsive to current traffic condition, and promotes continuous improvement.

What is most important, is it is easy for a user to understand to improve continuously.

Damage and Cleaning Image Recognition for Containers

Just finished a course on Machine Learning on Coursera and found that we can train a machine to recognize letters. Much like how Optical Character Recognition is developed. By the same approach, one can do the same, and have already done, for face recognition for images. With some many containers passing through Singapore, and with so many pictures taken of damaged containers for repairs, we can actually use this to train a machine to recognize damaged containers for areas of repair and its cost. One can certainly use current CCTVs or other types of video cameras to capture photos. 

The manpower and processing needed currently to do this will be saved.

Improve Forecast To Lower Safety Stock

We have all been taught the safety stock formula. It is however, not very responsive to use average sales and historical sales to calculate the sigma value. This is because the sales changes with time, and we are all unsure of the period of historical sales to average it over. Additionally, new product launches, promotions and variants all affect actual sales numbers. 

Since the sigma value for safety stock is a measure of the difference between projected vs actual sales, it makes a lot of sense that sigma be derived from forecast vs actual sales. This is a more responsive and forward looking approach as forecast incorporates up-to-date factors like sales, promotions or events. Plus, we are using the actual sales rather than average sales to derive the sigma for safety stock. This has significant implication for re-order point calculation subsequently.

What's more, measuring forecast and actual sales also facilitates Sales & Operations Planning in an organization.