COORDINATING HUNDREDS OF COOPERATIVE, AUTONOMOUS VEHICLES IN A WAREHOUSE Dev Bahadur Poudel Automation and Control Seminar Jacobs University Bremen Monday, July 30, 2012 Automation and Control Seminar 1
COORDINATING HUNDREDS
OF COOPERATIVE,
AUTONOMOUS VEHICLES IN
A WAREHOUSE
Dev Bahadur Poudel
Automation and Control Seminar
Jacobs University
Bremen
Monday, July
30, 2012
Automation and Control Seminar 1
Structure of Presentation
Monday, July 30, 2012Automation and Control Seminar
2
Overview on Multi-Agent System
Multi Agent System in Warehouse
Kiva System in Warehouse
Autonomous Robot
Monday, July 30, 2012Automation and Control Seminar
3
Robots that can perform desired task in
unstructured environment without human
guidance.
Have some degree of autonomy
Have control over their internal state and their
behavior
Agents
Monday, July 30, 2012Automation and Control Seminar
4
Computer system that is capable of
independent(autonomous) action to satisfy
design objectives
Problem Solving entities
Situated in a particular environment (input
environment through sensors and act on
environment through effectors)
Multi-Agent System
Monday, July 30, 2012Automation and Control Seminar
5
Agent Oriented Approach to solve Complex
Problem
Consist of a number of agents interacting with
each other to accomplish the goal in a
common environment
Problem are decentralized to different agents
Decomposition of Knowledge and Expertise
Distributed Intelligence
Why and Where Multi-agents
Systems?
Monday, July 30, 2012Automation and Control Seminar
6
Decentralization of Problems and Knowledge
For Complex Control Systems
Increase the productivity and decrease the cost
Reliability, Robust
Military Network-centric Operations
Search and Rescue
Transportation and Logistics
Now days used in Warehouse to increase the efficiency
Warehouse Scenario
Monday, July 30, 2012Automation and Control Seminar
7
Goods-to- Customer
Traditional Automation in
Warehouse
Monday, July 30, 2012Automation and Control Seminar
8
Pickers(human) move around the Warehouse
Fetch products and return them to packing
station
Human have to look for the order items
Order items in Conveyors
Batch Processing
Time consuming
Modern Automation in
Warehouse
Monday, July 30, 2012Automation and Control Seminar
9
Orders’ item come in hand of picker like a
MAGIC!
Use of Autonomous agents
Agents search the order fetch to the human
Agents co-ordinate to achieve a system goal
Many orders can be fulfilled
Increases Productivity
Kiva System
Kiva System
Monday, July 30, 2012Automation and Control Seminar
10
State-of-the art for modern automation in
warehouse
Founded on better approach for order fulfillment
(goods to customers)
Uses Hundreds of autonomous mobile robots
Sophisticated control Software
Uses the concept of Distributed Intelligence
Founded in 2003 (Mick Mountz )
Fielded in 2006
Implements Distributed Intelligence
Resources
Monday, July 30, 2012Automation and Control Seminar
11
Inventory ( with dimensions and frequency)
Inventory Pods (pods can visit one or multiple
stations)
Bins (pods can have 1-1000 bins)
Order Pods
Parking space for pods
Robots
Picking Stations
Kiva System Layout
Monday, July 30, 2012Automation and Control Seminar
12
Green: Storage Area
Orange: Mobile Robots
Blue: Stations
MAS Architecture in
Warehouse(Kiva)
Monday, July 30, 2012Automation and Control Seminar
13
Job Manager
Inventory Station Drive Unit
Agent 1
Agent 2Agent 3
Drive Unit Agent
Monday, July 30, 2012Automation and Control Seminar
14
Mobile Robots
Fetch the Inventory pod to the picking station
Take the Order pods to the shipping station
Transport the Inventory Pod to replenishment
Path Planning
Inventory Station Agent(ISA)
Monday, July 30, 2012Automation and Control Seminar
15
Picking stations : Workers pick items
Replenishment stations: barcodes are
scanned and appropriate pod come to the
station
Report accomplishment of its tasks.
Equipped with computer that pick lights,
barcode scanners, laser pointers used to
identify the pick and put locations
Job Manager
Monday, July 30, 2012Automation and Control Seminar
16
Central Server System
Resource Allocation
Communicate with Warehouse Management
System
Receives customer orders that need to be
fulfilled
Assigns drives, pods, and stations to carry out
the tasks.
Kiva Robots and Pods
Monday, July 30, 2012Automation and Control Seminar
17
Kiva working Mechanism
Monday, July 30, 2012Automation and Control Seminar
18
JM receives orders and assign to stations to
fulfill
Robot(Drive Unit) carries the inventory pods
Inventory pods come to picking stations
Picking: Workers pick items and put in order
pod
Shipping : Order pod move for shipping after
fulfilling all the orders
Replenishment : Inventory pods go for
replenishment
Pick-Pack-Ship
Order fetch Configuration
Monday, July 30, 2012Automation and Control Seminar
19
Path Planning
Monday, July 30, 2012Automation and Control Seminar
20
Optimization of the path to fetch order from
storage to the station
A* Algorithm
Travelling Salesman Problem
Resource Allocation Challenges
Monday, July 30, 2012Automation and Control Seminar
21
Objective: Optimizing the system(keep the
workers busy minimizing the robots and pods
used)
Which order to assign to which robot?
Which pod to pick up?
Where to keep the pod after order is fulfilled?
Which pod to send for replenishment?
Optimization(make all the robots and stations
busy)
Heuristic Technique for
Optimization
Monday, July 30, 2012Automation and Control Seminar
22
Order Allocation
1.Time to fulfill the order must be minimized
2. Inventories around the station and in
queue pods must be considered
Inventory Pod Selection
1.Nearer pods
2. Multiple items to be picked in one visit
Resource Allocation
Monday, July 30, 2012Automation and Control Seminar
23
Pod Storage Allocation:
1.Slow frequency pods kept backward
2.high frequency pods are kept nearer
Robot Allocation:
1.Pick More pods using less Robots
2.Decrease the Queue in stations
Replenishment Allocation:
1.Maximize the cubic utilization of the pods
2.Bin packing Problem
3.Create faster pods and slower pods
Advantages
Monday, July 30, 2012Automation and Control Seminar
24
Increase Productivity: double the output
Lower Cost: lower installation and operational
costs than traditional warehouse automation
systems
Location free replenishment : items can be
kept in any pods
Adaptive storage: Pods are store using
heuristic
Expandability :Add inventory pods and drive
units to increase throughput during peak
season
Challenges
Monday, July 30, 2012Automation and Control Seminar
25
To the Computer Scientist
Development of an appropriate high level software infrastructure/framework to support the building of multi-agent systems
Efficient Algorithm for Optimization
Path Planning for robots
Resource Allocation
Coordinating Robots
Dynamic, Stochastic and hence Intractable
Kiva in Action
Monday, July 30, 2012Automation and Control Seminar
26
Have a Nice Day!
Monday, July 30, 2012Automation and Control Seminar
27