Rise of the logistics robots
Robot control systems have progressed from simple relay logic to highly sophisticated dedicated controllers. Current system algorithms for the more capable autonomous robots require the latest processor capability and often require custom computer system design. With the improved computational capabilities come additional robot capabilities, including completely autonomous operation in complex, dynamic environments.
Robots are available as stationary or mobile, and in multiple form factors depending on the job at hand:
- Fixed-in-place robotic arms
- Automated guided vehicles (AGVs) following fixed pathways
- Autonomous intelligent vehicles (AIVs) able to freely roam a defined space
Robots lower labor costs and can improve quality with repetitive operations. Pick and place robot arms are used for such mundane tasks as removing bread trays from a continuous‐belt commercial oven, dumping the baked loaves onto a conveyor, and stacking the empty trays. Apple has a robotic disassembly line that takes iPhones apart to recycle the components and recover precious metals. This is brainless, numbing work and a robot is perfect in place of a person.
What is common to robotic arms is they are fixed in place and operate in a defined and limited space. They are typically controlled with a dedicated controller or programmable logic controller with a defined set of operations. As they are not spatially aware, a fence or other protection usually surrounds them to prevent people from getting injured. They require the parts they handle to be located in a certain place with a specific orientation, and what they assemble will also be in a defined location.
Vision systems and controllers have progressed to the point that assembly robots can see the environment and work piece orientation, and they can adjust their motions accordingly to capture these parts. Instead of only capturing parts in the same orientation in the same location, robots can now scan a bin of parts, pick one out, and correctly orient it for attachment. But even with vision, they are fixed and not mobile.
Automated guided vehicles
Delivery robots are an early form of logistics robot and are considered an AGV. An example would be mail delivery robots, which are currently in use. These robots move correspondence and packages around offices and other large, protected areas and can be configured to use elevators for multi‐floor routes. They depend on a glide path embedded in the floor – typically a wire, colored line, or magnetic strip – for navigation and they follow the marker strip though the hallways on a fixed, predetermined route. Sensors such as ultrasonic transducers and pressure-sensitive bumpers allow the robots to avoid collision with people or objects left in their path.
Industrial AGVs come in all forms and sizes and are commonly employed to move raw material from the receiving area to the warehouse or from the warehouse to the production areas, and move finished goods to the warehouse or to the shipping department. They can be configured to handle and move packaged material and can include a built‐in fork lift to stack material on elevated racks or just about any other material handling technology, such as side clamps for cartons not on pallets. Tugger AGVs, which tow carts using an actuated hitch, can even haul multiple trailers for additional capacity.
There exist hybrid forklift AGVs, which can operate autonomously or with an operator sitting in the vehicle. They can be used for work‐in‐progress for the repetitive movement of material within the manufacturing environment from station to station. AGVs are commonly used in the pharmaceutical industry as the system tracks vehicle movement for process validation. A relatively new application is loading semi‐trailers.
On the larger side, ECT in the Netherlands has introduced automated sea‐going container movement at the Port of Rotterdam. At Rio Tinto’s Pilbara mines, the operator has mapped the entire open‐pit mine environment and is using 22 driverless mine trucks (huge multi‐hundred ton dump trucks) to haul all the iron ore from the pit. Supervisory control is handled in Perth, Australia, 1,200 miles from the mine. To the naked eye, the operation looks like a normal mine, but the trucks operate 24 hours a day, 365 days a year with no risk of accident from operator fatigue.
AGVs at Amazon
Kiva Systems, now Amazon Robotics LLC, developed a semi‐autonomous delivery robot that Amazon put to good use in their fulfillment centers.
Prior to the use of Kiva robots, people called "pickers" picked orders in Amazon’s huge fulfillment centers. One warehouse in Baltimore, for example, covers one million square feet. The pickers there were responsible for up to 120 picks per hour and would walk (or run) up to 13 miles per day during a 10-hour shift. Amazon is replacing these pickers with movable racks and Kiva robots.
A typical store stocks its shelves in a logically sorted manner so customers can easily find the items they want to purchase, but that approach would not work efficiently for Amazon. Amazon receives shipments and those items are stored in any empty bin on any one of the racks. A single bin might contain disparate items not related to each other. The warehouse management computer registers that item location for later retrieval. Pickers once carried a scanner that told them where in the warehouse an item was stored. They would retrieve the item from the rack and place it in a bin on their cart, then move to the next item location. A rack location could hold different items and the picker would choose the correct item as directed by the scanner.
Since the implementation of the Kiva robot, the picking process has changed. The Kiva robot is designed to move under freestanding, unanchored racks and lift them off the floor. When an order is entered into the warehouse database system, the nearest available robot moves the rack with the item(s) to be shipped to the picking/packaging station where a person retrieves the product from the rack, boxes it, and sends it on its way. The rack would then be moved back to the warehouse area or to the receiving area for the empty bins to be stocked. The rack does not have to go back to the original location. Kiva offers two models: one can lift 1,000 pounds and another can lift 3,000 pounds. Every hour, each robot’s onboard battery requires a recharge, which takes about 5 minutes.
Kiva robots keep track of their location by scanning bar code stickers on the floor and operate in a defined grid pattern. The supervisory computer keeps track of the location of all the robots moving around the warehouse so they don’t run into each other and are moved to where they are needed. The robots also communicate with each other to avoid collisions. These robots appear to operate in barely controlled chaos (Figure 1).
While the Kiva robots are semi‐autonomous, they follow defined paths using the barcodes on the floor to tell them where they are. They are not spatially aware. For safety purposes, Kiva robots can only work in areas that are cordoned off from people.
Intelligent, free roaming helpers
Adept, a division of Omron, manufactures the Lynx aiv (AIV). The Lynx vehicles, unlike AGVs, are spatially aware and highly intelligent with the capability to operate in a complex, changing environment with no guidance from a centralized computer.
Adept manufactures an intelligent mobile base, relying on their customers to incorporate material handling mechanisms attached to the top of the robot. The base unit provides local control, collision avoidance, communications, and mobility. Customers have created a wide variety of attachments, including racks for slide‐in trays, conveyors to accept packages from fixed conveyors, and even multi‐jointed arms with applications ranging from logistics to light‐duty manufacturing, semiconductor manufacturing, line‐side replenishment, hospitality, and healthcare.
Adept provides a fleet management system called the Enterprise Manager 1100, manufactured by Chassis Plans. The Enterprise Manager is the bridge between the entire mobile fleet and the rest of the facility, providing centralized status monitoring, job dispatch, configuration management, IO‐handshaking, and integration with other entities throughout the facility (Figure 2).
[Figure 2 | The Enterprise Manager 1100 is rack mounted in a 1U form factor and provides a bypass NIC that allows two units to be cascaded together, providing continuous real-‐time backups and quick recovery in the event of hardware failure.]
An example installation is in a fulfillment center moving boxes from 28 conveyors to 10 shipping stations. Conveyor utilization changes with the shopping seasons, so not all conveyors are used throughout the year. The same applies to the shipping stations. Initial analysis of the conveyor buildout indicated the entire space would have been filled with conveyors (Figure 3). Instead, the 28 conveyors terminate in individual transfer stations that can notify the Enterprise Manager when a package is ready, shift the package to a conveyor on top of the Lynx, which then can move the package to an available packing station as directed by the Enterprise Manager.
The Lynx AIV is spatially aware of its surrounding environment – it knows where it is by looking around and consulting an internal map. For initial setup, a Lynx is guided around the working area, either by using a hand‐held controller to steer it or by following a person walking the hallways or open areas. The Lynx uses its on‐board sensors to “see” the environment and build an internal map. The operator then has the ability to edit the map marking areas that are off limits, areas with reduced speed limits, termination points, and so forth. The map only has to be created once and other Lynx robots in the system can use the same internal map.
A key feature of the Lynx AIV is independent dynamic path planning. These robots are used to move material from one point to another. In a dynamic environment where pallets may be stacked between the robot and its destination or a person may be standing in a doorway on the shortest path to the destination, the Lynx can reference its internal map to autonomously plot an alternate route to the destination. The sensors used for navigation also allow the Lynx to safely share the workspace with people or material stacked on the floor. If a person is standing in the way, the Lynx can simply shift its path around the person if there is room, find an alternative path, or stop if there is danger of collision.
The mechanical components comprising mobile robots are straightforward. Motors, wheels, gears, and motor controllers don’t change much over time. On the other hand, controllers and management appliances, which are based on modern processors, allow the use of much more sophisticated algorithms and sensor fusion. This added processing capability and spatial awareness through more sophisticated sensors will allow increasingly more human‐like behavior with truly independent action in complex, changing environments.