The client manufactures elevators for the urban skyscrapers and has been reinventing the outlook towards elevator and escalator transportation for more than a century.
One of their innovative products aims to achieve intelligent transportation, enhanced passenger convenience, and efficient power utilization. The product integrates with the infrastructure of existing elevators and calculates the best possible ways to transport passengers based on factors like usage traffic patterns, weather, train delays, etc.
The client chose to evaluate GCP’s Machine Learning components to achieve intelligent transportation. Their major objectives were to reduce the time for which a passenger waited for an elevator and to make elevator operations more power efficient. We designed a GCP-based Machine Learning Minimum Viable Model (MVM) using TensorFlow on Google Cloud AI Platform and BigQuery.
The engagement was carried out and delivered in three steps:
- Load Data into BigQuery: MediaAgility received CSV files from the client and loaded them to Google BigQuery.
- Develop and Train ML Model: MediaAgility trained a Machine Learning model using Google Cloud AI Platform and TensorFlow.
- Knowledge Transfer: MediaAgility conducted a knowledge transfer session for the client’s team to explain ways to further improve the model.
The Minimum Viable Model identifies the expected usage patterns and predicts on which floor should an idle elevator be parked at different times of the day to lower the time to wait for passengers. Further, with the usage patterns, the client decides when to put the elevators in standby mode for reduced power consumption.
This engagement is a chronicle of the extent of ML’s capabilities and impact; ML isn’t just for the high-end challenges, but it is also for the day-to-day operations which are hidden bottlenecks to business. The client improved their passenger experience and power efficiency – something, seemingly, not influencing the business directly, but nonetheless powerful transformations and efficiencies – a person having to wait for lift as he is running late for a meeting can testifyArpit Agrawal, Google Business Unit, Persistent