What is Warehouse Automation
March 18th, 2021
Warehouses play a very vital role in the 21st Century. Today, most of the E-Commerce companies want to ship orders as fast as possible to increase customer satisfaction. Some companies also have guaranteed 1-day delivery commitment but this is only possible when the goods from the warehouse are retrieved and sorted for dispatch very fast. For large warehouses, hiring people for performing such activities is not a good decision, instead now companies are going for Automation. There are various automation solutions for warehouses but there is no solution available, which can individually pick and place orders from warehouse racks to receiving or dispatching counters.
Technology has changed the traditional warehouse management system by altering the way companies send and receive orders. Traditional systems can be problematic because as a company’s volume order increases, the system of receiving and distributing orders becomes more complex. Companies often expand product offerings, meaning warehouse management systems must adapt to new inventory and the changing business. A poor warehouse management system drains a company’s money by not accurately keeping track of orders.
The order picking or order preparation operation is one of a logistic warehouse's processes. It consists in taking and collecting articles in a specified quantity before shipment to satisfy customers' orders. It is a basic warehousing process and has an important influence on supply chain's productivity. This makes order picking one of the most controlled logistic processes. It is one of the warehouse management system functionalities.
To achieve lower warehousing cost, accurate and faster order fulfilment, higher inventory traceability, intelligent automation systems must be used. One such automated system has been utilized by Grey Orange, a smart butler as shown in the figure above, which can pick-up entire stations with multiple storage units and can move at a relatively high speed. One major disadvantage of this smart butler is that it cannot carry individual units from one place to another without carrying the entire station. This is somewhat inefficient in terms of energy spent to carry the entire station and back in cases where only a few products are required.
Another example of intelligent automated robots is the Pick-up robot by Magazino. It is a smart bot which picks up the inventory required and places it in the shelves built within it. When its shelves are filled, it goes to the unloading station.
One major drawback of this design is the limited number of shelves which limit the efficiency of the bot. The bot has to unload the inventory at the unloading station at regular intervals of time, therefore this design cannot be used in a large warehouse where there would be dozens of orders to be collected simultaneously.
After carefully studying and analyzing the above-mentioned examples we have come up with a unique design for a smart robotic warehouse solution to cope up with high demand in a minimum period of time. Our proposed design consists of two bots - Master bot and Slave bot. The Master bot as the name suggests would be the one collecting the inventory from the warehouse racks. The Slave bot will have shelves built on it and will be following the Master bot. The Master bot would collect a single unit at a time and place it in the slave bot following it. The master bot is designed to carry a maximum load of 50kg at once and can rotate a complete 360 degrees. The Master bot will load each item into the Slave bot and the Slave bot will keep following the Master bot. Once the capacity of the Slave bot is full, it leaves for unloading and another slave bot comes and takes its place following the Master bot. The Master and Slave bots would be connected using swarm robotics where the Master bot would be the main control system and the Slave bots would follow it.
This Master-Slave concept will make the entire order retrieval process faster and efficient. There will be no delay in between collecting huge orders. Both the Master and Slave BOTs work autonomously. There is a mainframe system that obtains the order and retrieves the information about the position of the items in the order. The mainframe passes this information to the Master BOT.
The Master BOT has a 2D map of the warehouse and has information about the position of each item to be picked. Then the Master BOT routes itself to pick up each item with a slave bot following it. In the below figure is an example of how Master BOT routes itself.
The Slave BOT is following the Master BOT by using a 2D map location of the Master BOT and also distance sensors once it has reached behind the Master BOT. Once the shelves of the Slave BOT it full, it autonomously goes to the unloading area and another Slave BOT goes to the Master BOT’s location and follows the Master BOT.
Both the Master BOT and Slave BOT was designed to withstand high loads and move at higher speeds even with load. The movement of the BOTs is facilitated by a differential drive using two high torque Delta servo motors. The servo motors were controlled by a PLC through CAN bus protocol. Both BOTs have an HMI Display to view status and also control them manually.
This is the base design of the BOTs. There are multiple components enabling the control of the BOT. A PLC (Programmable Logic Controller) acts as the brain of the BOT and. A HMI Display is used for viewing status of BOT and also for manual control if necessary. The BOT has two Delta ASDA-A2 servo motors and drives for movement.
The Master BOT has a forklift controlled by a Linear Motor (LU Series - Delta). The forklift can move forward and backward to pick up the pellets of the item in the order. The forklift also moves up and down to align with the shelf or row to pick the item from using a lead screw mechanism. The lead screw mechanism (ball screw mechanism can also be used) used two servo motors for up and down control of the forklift.
The mainframe system also has a PLC and a HMI. The order to be collected is fed to the mainframe PLC. The mainframe PLC divides the order to each Master BOT and shares the location of each item to be picked to respective Master BOT. To enable, sharing of such data we have used Wi-Fi modules on the Master BOTs, Slave BOTs and the mainframe. Each Wi-Fi Module is connected to the respective PLC and a set of registers is defined as a shared resource. Any change made in these registers from one PLC will reflect into all the other PLCs through the Wi-Fi module. This mechanism is used by the bots to communicate with each other and also with the mainframe.
Given Below is a Concept Design for the proposed Warehouse system. The Master BOTs take products from the shelves and load on the Slave BOTs and the Slave BOTs load the products in the unloading area once capacity is full. This concept solves delays involved with retrieving huge orders.
Another main issue in Warehouse Automation is dividing order to multiple Master BOTs. In any warehouse due to the size of the warehouse there will be multiple Master BOTs. Traditionally each Master BOT is assigned a specific region and collects items of the order from that region. The allocation of Master BOT region is done in the beginning only. Once a set of orders arrives, the job is distributed among the master bots based on the region from where it has to be collected. The disadvantage with this approach is that the bot which has fewer items to collect will finish the task earlier and will remain idle.
Let us consider an example to demonstrate this. In a warehouse there are six shelves and three Master BOTs. Traditionally each Master BOT will be assigned a particular region to operate in. Since there are three Master BOTs, each Master bot is pre-assigned 2 shelves each.
In this example, the first Master BOT has to collect 24 items, the second Master BOT has to collect 10 items and the third Master BOT has to collect 16 items. This distribution is not equal. The second Master BOT will finish the collection of its items and remain idle for a long time till the first Master BOT finishes collection of its items. The order can only be further processed once all the Master BOTs are done with picking up the items.
This problem can be solved by dynamically distributing the task between the orders. The Master BOTs do not have a specific region allocated in the beginning. The region for each Master BOT is dynamically allocated once the order to be collected is received. The allocation happens in such a way that all the BOTs have equal workload so that the entire order retrieval process is efficient.
Using the Concepts Machine Learning, the workload is dynamically distributed among master bots. The location of the items in the order is considered as points in a 2D space and then a clustering operation is performed to divide the points into n groups. Here, n is the number of Master BOTs. The clustering algorithm used is Fuzzy c-means clustering. This clustering algorithm divides the order equally to all the Master BOTs.
OPTIMISATION USING MACHINE LEARNING
preassigned areas. To Optimize the warehousing solution, we have used Machine Learning
Scenario: A warehouse with 108 products shelved in 12 two sided shelves.
Number of Bots: 4 (Master bots)
Simulation: An order of 100 products (randomly generated)
This is how traditionally the order would be picked up by the Master BOTs. Each Master BOT is assigned a quarter in the beginning only and collects items in their respective regions.
Using clustering, the order allocation to each Master BOT can be done better. In the below image it can be seen that the order is almost equally distributed between the Master BOTs. This is an example of dynamic distribution of workload between the Master BOTs to enable efficient order retrieval.
Scenario: A warehouse with 30 products shelved in 6 two sided shelves.
Number of Bots: 3 (Master bots)
Simulation: An order of 50 products (randomly generated)
Traditional Order Allocation to Master BOTs
Optimised Order Allocation to Master BOTs
In both the cases the clustering approach to allocate items to pick up to Master BOTs is more efficient than the traditional approach. This concept makes the entire warehouse automation process efficient and convenient. Both the Master-Slave BOT concept (Swarm robotics) and the dynamic allocation of workload between Master BOTs (Clustering - Machine Learning) helps to optimise Warehouse Automation.
- Orders are obtained
- Once the buffer for orders is reached, the set of orders are given to the BOTs.
- The job of collecting the orders is dynamically divided among the master BOTs using Machine Learning (Clustering).
- Each master BOT gets its own set of order to fetch.
- A Slave BOT is allocated to each Master BOT and the Slave BOT follows the respective Master BOT to store the collected item.
- The Master BOT goes to the respective shelves and collects and places items in the shelves of the Slave BOT using the forklift.
- Once the Slave BOT’s capacity is full it goes to the unloading area while another Slave BOT is deployed to follow its Master BOT.
- Once the task is completed, the next buffer of orders is taken for execution.
- We can modify the Master bot to increase its capacity of carrying load and the limit of the slave bots can be regulated based on the industry it is being used.
- Using the clustering algorithm used in the bots, these bots can be used in very large warehouses to reduce the total time taken to collect the orders.
- It can be implemented in large libraries and supermarkets for quick response and better efficiency.
- Complete automation of warehouses using swarm robotics would result in time management of demands, as the bots are scalable, reliable and flexible.
ABOUT THE AUTHOR
Cyril Joe Baby is an Entrepreneur, Developer and Researcher. Co-founder and Chief Technical Officer of Fupro Innovation Private Limited, a high-tech research company in prosthetic and rehabilitation devices.
Cyril has worked on various projects and research works during the past years. (Google Scholar link) His fields of interest and expertise include Artificial Intelligence, Robotics and Embedded Systems. Cyril continues to contribute to these fields by working on various exciting projects and guiding beginners to work on their projects. For any further information, you can contact him at email@example.com. firstname.lastname@example.org/
It started off in a more hectic manner than I could expect. ... read more
- Priyanshi Somani, Manipal Institute of Technology
“GAIP is perfectly aligned with someone's goal who wishes to experience an outburst of academic challenges while working on projec ... read more
- Sukriti Shaw, SRM Institute of Science and Technology
“Combining different characters and skillset from different institutes and domains in a new country and fantastic institute, it wa ... read more
- Shaolin Kataria, VIT, Vellore
“An enriching and enthralling experience. The course was extensive but worth every penny. ... read more
- Arudhra Narasimhan V, SASTRA DEEMED TO BE UNIVERSITY
“I personally learned quite a bit here but the 6-month project or LOR aren't as easy to get as was portrayed before. ... read more
- Dwait Bhatt, BITS PILANI
“It was a great experience for me, and far beyond my expectations. ... read more
- Shrikant Tarwani, LNM Institute of Information Technology
“This Internship is the perfect balance of theory and practical application. ... read more
- Mahima Borah, Manipal Institute of Technology
“This Internship has strengthened my concepts on Artificial Intelligence and Deep learning which are the hot words of today’s t ... read more
- Mansi Agarwal, Delhi Technological University