Yuanhong Zeng
Muhan Zhang
Work under the supervision of Professor Ankur Mehta
This Final Design Review presents the development of an advanced warehouse pick-and-place system designed to enhance efficiency and reduce costs associated with manual picking processes in warehouses. The primary focus is to implement a system capable of accurately identifying and accessing a specific 2D location to retrieve items from uniquely assigned bins and securely transfer them to another location.
The current manual processes in warehouses are costly and error-prone, with a significant percentage of e-commerce returns attributed to picking errors. This project aims to address these challenges by introducing an Autonomous Mobile Manipulator (AMM), which combines mobile robotics and advanced manipulation capabilities to automate picking tasks.
The system is designed to operate within predefined warehouse environments, navigating through aisles with precision and interacting with bins placed on standard shelving units. It utilizes a combination of sensors, including LiDAR and RGB-D cameras, to navigate and identify items, ensuring high accuracy and minimal picking errors compared to human pickers.
The design review covers the technological aspects, problem definitions, subsystem design, and preliminary analysis conducted to evaluate the system's feasibility and performance. The system's architecture is outlined in detail, highlighting the integration of physical and cyber layers to facilitate seamless automation in a warehouse setting.
The project outcomes indicate that the proposed system can significantly improve picking accuracy and efficiency, with potential implications for cost savings and enhanced operational throughput in warehouse management.
Most warehouses today still maintain a manual picking process while picking consists more than 50% of their operating costs (Gajšek et al., 2017). However, there is a 0.26% average picking error rate of manual picking (wrong item, quantity, or omitted order) (Rammelmeier et al., 2011), and 23% e-commerce returns are because of wrong item received (Saleh, 2023).
In addition high demand for efficiency and throughput worsens the working condition for inventory workers. Workers must pick an item for packing every 30 seconds during their 10-hour shifts, with their tally clearly displayed. Others must pack 120 items per hour, with supervisors rating their performance and providing reviews. Sitting is strongly discouraged, bathroom breaks are timed (Zhang et al., 2021). In addition the repetitive nature of the industry—lifting, straining, bending and twisting-- leads to a wide variety of musculoskeletal injuries involving the neck, back, knees and shoulders.
In today's fast-paced production environments, businesses are constantly seeking ways to accelerate processes and enhance efficiency in pick and place operations.
Previous attempts in improving warehouse pick and place solutions can be classified into two categories.
Manual Solutions: This encompasses "Pick by Light and Put to Light" systems , which utilize color-coded light signals to direct pickers based on order information. In RF Picking, operators use RF scanners to scan item barcodes, with the device confirming the accuracy of the selection. Voice Picking operates similarly; however, workers verbally confirm item IDs, and the system verifies the correct item has been selected through voice recognition, providing feedback via headphones. These solutions greatly improves the accuracy of picking to 99.9%, however, they are still limited by the efficiency (movement speed and picks-per-minute) of human pickers (Rammelmeier, 2011).
Automatic Solutions: By automating the picking process, warehouses can handle higher volumes of orders with greater speed and accuracy. One of the key technological advancement is the warehouse management software (WMS). It is integrated with various hardware components, and it can optimize picking routes, allocate resources efficiently, and enable seamless coordination between different areas of the warehouse. In addition Autonomous Mobile Robotics (AMRs) and Manipulators equipped with advanced sensors and algorithms can navigate warehouse aisles, pick items from shelves, and transport them to designated locations. These solution optimizes throughput and accuracy, however they requires initial investments over $1 million (Hvilshøj et al., 2012), and they are inflexible with respect to throughput and Stock Keeping Units. This made them unsuitable for small business owners.