Vision

The Agilis robot uses a proprietary vision system that combines leading edge technologies for simulating stereo vision with structured light techniques. This vision system coupled with an array of positional and edge detecting sensors, allows the robot to “see” its work environment.

In conventional palletizing and depalletizing solutions and in typical storage and retrieval automation systems the user fixtures the product so that it is in a well-defined position relative to the robot. In the Agilis solution pallets are stored in conventional racking as is typical in the bulk storage areas of a distribution center, with no fixturing or constraints – thus imposing less restrictions and infrastructure costs on the customer. The vision system allows the robot to discover the exact positions of entire pallets, the nesting pattern of cases on the pallet, and the positions of each individual case on the pallet.

Furthermore, the active vision system coupled with the other sensors allows the robot to make on-the-fly adjustments such as when a case has been moved from its original position (e.g. somebody comes along and bumps the case) or when a case is missing (e.g. somebody has removed a case to fill an order elsewhere in the warehouse). This is one example of how the robot is able to use sensory data coupled with its built-in intelligence to adapt to a changing work environment.

Motion

The Agilis robot is a highly modular automation system comprised of a number of mechanical innovations. The base unit rides along a track that is laid at floor level in an aisle between two rows of conventional racking. By its’ modular nature the track’s length can be set to suit the needs of the customer, though the typical installation uses a track length on the order of 300 feet. The overall footprint of the system is much smaller than the equivalent footprint for manual operations on the same set of products.

The base unit carries with it a vertical mast, another system which is highly modular and can be built to suit warehouses of varying heights. A carriage system rides upon the mast allowing the robotic arm to traverse in the vertical direction, thus giving access to the multiple levels in a typical racking system (racking can consist of 4 or even more levels depending on the overall ceiling height of a warehouse).

Typical industrial robotic arms are comprised of a number of fixed-length links connected via revolute joints. One limitation of this design is in gaining access with the end-effector to tightly constrained spaces. In the Agilis solution we have sought to minimize overhead clearance space above a pallet sitting in the racking system, so as to allow maximum density of the racking system in the vertical direction. Therefore, our robot needs to fit into as tight a space as possible, while maintaining a high degree of mobility so as to be able to reach cases on both sides of the aisle. To this end we have designed a patent-pending telescoping robotic arm which is capable of fitting in extremely tight spaces while still reaching points more than 6 feet away.

The Agilis gripper combines power and novel sensor integration, allowing it pick up cases weighing over 60 lbs and feed the intelligence system information about case position and state.

Intelligence

Most industrial automation solutions today are directed by pre-programmed directives that do not accommodate changing conditions or unstructured environments. The robots are taught a preplanned path and execute that path repeatedly without regard for external influences such as a change in work piece position. In contrast, at the heart of Agilis is a true intelligence system that performs its own planning and is capable of adapting to its work environment.

The first exhibition of this intelligence is in the robotic system’s capability to discover product on its own. Using the proprietary vision system and complementary sensors Agilis is able to discern the position, size, case make-up, and nesting patterns of any pallet inserted into its workspace. From this data Agilis can build a dynamic model of the position and state of every single product in its workspace: a model which it continually updates as it continues to collect data during picking activities.

During a picking shift Agilis is sent data from the customer on cases that are to be picked. As the typical line consists of two robots with overlapping workzones one of the first steps is for Agilis to intelligently distribute picks across the two robots in order to maximize picking efficiency. In the case where a warehouse has multiple robotic lines and identical product is present in more than one work cell, this optimization will be balanced out across all lines. Each robot computes its own motion plan and picking order to minimize pick time. In the case where a robot runs into a problem acquiring a product or is lagging behind in time, picks will be redistributed and rebalanced dynamically across other robots and other lines.

Agilis also directs its own replenishment activities as cases are depleted from pallets. In doing so the system seeks to insure no interruption of picking activities due to unavailability of a product. Agilis also optimizes replenishment activities to position new pallets in empty slots that will minimize future pick times, based on historical trends and known pick orders.