Some suggestions for improvement
of the current Kiva system
Vrije Universiteit, Artiﬁcial Intelligence Department,
De Boelelaan 1081a, 1081 HV Amsterdam, The Netherlands,
Course: Bachelor Referaat , supervisor: Zhisheng Huang
Abstract. Kiva has improved overall performance of distribution centers by implementing robotics and concepts from Artiﬁcial Intelligence. Although the results look promising, there is still some room for improvement. Small changes in the environment or robot behavior rules might improve performance once more. On the other hand results of fundamental changes like introducing chaos are much more uncertain. This paper sketches some changes that could improve the performance of the current Kiva systems.
This paper starts with a short introduction on the Kiva system, a deﬁnition of some basic concepts and an explanation of control methods. After the introduction some changes to the Kiva system are given. Readers should have an interest in automated warehouses and robotics. Basic knowledge of computer systems might be useful.
The Kiva system [9, 2] is a solution for automating pick, pack and ship orders used in distribution centers. It takes over product search and product retrieval assignments from human workers. Instead of applying technology to ﬁt robots in an existing system, Kiva uses a custom environment.The ﬂoor of the distribution center is equipped with a barcode grid and small orange robots that ﬁt underneath special product racks. A central server sends a message to a particular robot to perform a product retrieval task. This message contains the coordinates on the grid, which are barcode-stickers and information which speciﬁc sub part of the rack contains the needed product. The robot is equipped with a barcode laser so it can read its current position. The robot then calculates a path by combining its current location with the product location mentioned in
Fig. 1: a) A Kiva robot b) a Kiva warehouse implementation c) a Kiva drive unit rotating underneath the lifting device  d) a Kiva station 
the received message and moves through the environment by making 90 degree turns on speciﬁc coordinates. When the robot arrives at the product location, the robot then lifts the product rack, which can contain multiple diﬀerent products, from the ground with a screw mechanism and transports the product rack to a station where a speciﬁc product is packed by a human worker to complete a customers order. When a robot arrives at a station the product space on the rack is highlighted with a laser so the human worker can quickly get the speciﬁc product. The human worker than scans the speciﬁc product and another laser highlights the box belonging to that speciﬁc order. After scanning the speciﬁc box the robot stores the product rack or uses the rack to fulﬁll another order at another location. Productivity is increased by saving search and retrieval time of products. It is more ﬂexible than existing solutions like “Automated Storage and Retrieval systems”, where products are retrieved by conveyor belt systems or autonomous forklifts placed on rails. The Kiva shows scalability when the
warehouse size increases only more robots need to be added instead of building a complete new system. Robustness is shown when a single robot breaks down. Only that speciﬁc robot stops working and not the complete system. Other non technical maintenance issues like cleaning the barcode readers on robots and replacing barcodes on the grid can be done quickly, while repairing and maintaining a conveyor belt system is a time consuming and expensive job. 1.2
Before addressing subjects like “Can the Kiva system be used for automated forklifts”, “Is Kiva a centralized or decentralized system” and “What can be improved in the current Kiva System”,...