—Introduction to the smart logistics project
Shanghai Jiuzhang Intelligent Technology Co., Ltd.
The Institute of Science and Technology for Brain-inspired
Intelligence (ISTBI), Fudan
Transport All Goods Easily
2
Contents
Project introduction
Development schedule2
1
3
Artificial
intelligence
Internet of
things
UnmannedBig data
 Logistics:Real world supply chain management
 AI logistics target levels and constraint levels
 AI logistics digitalization, platform
based,automation
 Project content and innovation points
 Project target exhibit
 Application cases in real world:How AI help
logistics suppliers realise better social targets
Project target:help realise intelligent logistics which
meets real world needs
4
Transportation strategy
Warehousing strategy
Network strategy
Dispatching
control
Forcast
group
Decision
making
planning
Cost
Output
Budgetary
constraints Institutional
constraints
AI
strategy
 Massive variables
 Multi-dimension
 Multi-size
 Dynamic
 Multi-target
 Strong constraint
 Non-formulization
 Non-Structured data
Maximize service & minimize cost under constraint
conditions for logistics optimization
5
AI helps grapple with the challenge of the future logistics
management development
Automation
Optimization
algorithm
Machine
Learning
 Unmanned warehouse
 Unmanned transportation
 Overall industries supply channel integration
 Real time and dynamic
 Non-structured logistics big data mining
 Forecast
6
Project characteristics and innovation points
Full supply chain
(B2B+B2C/road,rail,waterway)
Dynamic and real time
Non structural data graphical
representation
Brain-inspired intelligence new
algorithm new model experiment
platform
Based on current situation,
Look for future
(Artificial + intelligence )
Artificial intelligence 
Enhanced artificial intelligence
Optimization/enhanced learning
as the core
7
Classic case:Some large-scale automobile logistics
company project outline
• Production line
material
distribution
• After-sales
logistics
• Warehousing
planning
• Parts logistics
Warehousing Capacity
RouteScheduling
Data
Optimiz
ation
Study
Forecast
8
Classic case:Spare parts logistics transport optimization
learning algorithm
 Spare parts logistics transportation optimization needs overall
scheme :Transportation route optimization,loading optimization,and queuing
optimization.
 Overall optimization algorithm supports: 100,000 orders per day,with mileage
of 10 million KM.
9
Classic case:constraint conditions multi-target
combinational optimization
distrib
ution
distrib
ution
distrib
ution
Loading
diagram
Route
diagram
Loadi
ng
Queuing(Flow
control)
Unloa
ding
10
Classic case :difficulties
1. Large number of variables
2. More restrictions
3. Nonlinear , unformulated
4. Large solution space : >1062
algorithm
1. Accuracy
2. Adaptive, learnable strategy updates
3. Intelligent algorithm and business
adaptation coordination
Business
13 Loading constraint
8 Transfer library constraint
12 Loading constraint
8 Resource constraint
8 Time window constraint
14 Multiple rounds of transport constraints
constraints
Innovation points:Solutions to optimization
algorithm + learning algorithm
Intelligent Logistics Simulation
Module
Standardiz
ation
interface
Overall intelligent optimization
search algorithm
Analog
backout
Inheritance
algorithm
Taboo
search
Parallel
interfac
e
Machine learning
module
SVM Reinforceme
nt learning
1. W. Lu et al Intelligent Solution System towards
Parts Logistics Optimization ERUO2018 (Abstract,
Contributed Talk)
2. Y. Huang, B. Chen, W. Lu et al. Intelligent Solution
System towards Parts Logistics Optimization
WCGO2019 (Regular Paper Accepted)
12
Classic case: algorithms combination of sucessfully
researched and developed AI logistics
① SVM——Disassemble order strategy
② Greedy algorithm——Orders preprocessing
③ Community clustering——Initialization
④ Parallel analog backout algorithm
⑤ Self-adaption backout algorithm
⑥ Pruning searching algorithm with priority queue——Route
planning
⑦ Dynamic planning——Loading + route planning
⑧ Intensive learning,Q-Learning——Loading + route planning
⑨ Tree search algorithm——Post-optimization+route
combination
Generality
algorithm
development
Professionality
algorithm
research and
development
Classic case:the result better than some world
famous logistic management software
Result of a famous
business software
Our result
Load mileage 42235.5 39857.8
total mileage 84006.4
Dynamic loading rate 55% 52%
Number of vehicles 118 155
Number of cars 315
Running time 30分钟 10分钟
number of orders 2959 2959
The system under
algorithm has been at
online operation.
Classic case:Business oriented exact solution
15
Contents
Project Introduction
Development plan2
1
16
Business mode:Through technology innovation
cooperation,realise economic and social benefits
APP
Platform service
Vehicle & goods
match
Transport
management
AaaS
Algorithm:providing
customized algorithm
for logistics
enterprises
Algorithm
Scheduling
algorithm,route
optimization
Site selection
service ,Loading service
Economic benefits :
Reduce carbon emission
Reduce source consumption
Social benefits :
Realize cost-reducing & increase
benefits in whole society
Increase economic operation efficiency
Independent innovation:
Basic algorithm theory innovation
Basic software indepentent innovation
SaaS
Algorithm:providing
cloud/personalization
service mode
17
1、Develop a platform for optimization of
vehicle and goods matching:Goods which
can walk automaticly
• Freight transportation
platform like Didi or Uber
• Intelligently matched order
• Intelligently arranged route
and vehicle dispatch
• Route optimization
• Increase the income of the
freight carrier
• Reduce no-load rate
• Reduce carbon emission
• Reduce transactional cost
of vehicle and goods
matching
• Reduce loss and resource
waste
2、Develop SaaS System: Flexiblely embed existing
logistics management system
AaaS : jiuzhang VRP
Input:
Information of
vehicles
Orders info
Distance matrix
Output:
Scheduling
Optimized routine
3、AaaS(Algorithms as a Service)
Service of consultation plus algorithm provided
according to the clients’ business pain points.
Rrequirement survey
Interview research
Data survey
Establish a plan
Modeling
Adjust algorithm
Conduct tests
Select Range
Test summarisation
Promotion plan
Training and extension
Summarize and improve
starting
20
Future development and expectation
算法开发
场景模拟
应用验证
“Strong” Artificial
intelligence and brain-
inspired intelligence
Business→Mathe
matics→ Business
 Unmanned equipments
 Transport method
integration and
optimization(highway、
railway、air transport、
waterway transport)
 B2B and B2C integration
 Coordination trans legal
person and industries
(Connect data,platform
and equipments. )
21
Future overall optimization of logistic supply chain,
supporting the full load of a vehicle, express
transportation、spare parts, delivery,etc.
Solution Description
Route planning
Route optimization for multi-vehicle,multi-goods and multi-sites.Include two categories:1,Without
alternative loading and unloading.
2,With alternative loading and unloading(will load halfway)
Load planning
How to load the goods into the vehicle. 1,With the highest space use rate. 2,In the order of loading
and unloading.
Include:1,Loading schedule with the order of loading and unloading. 2, Loading schedule without the
order of loading and unloading.
3,Consider the center of gravity for the boxes transported in UAV.
Node location
Choose intermediate supply point(or the front warehouse).
In order that, 1,New network under logistic load is the most economical.(distance, cost)
Entry and exit queuing management
According to the vehicle entry time,unloading time,and with the help of the electronic fence,arrange the
order and time of vehicle entry and exit. Reduce the traffice jam at the gate of factories.
Three dimensional volume
calculation
With photo or surround video taking, 1,calculate the standard cuboidal object’s 3-dimensional volume. 2,
Determine the irregular objects’ outsourcing line.
Parts entering the factory
According to the spare parts’ location,route schedule and manufacturing plan,decide the transportation
plan and dynamic scheduling of the spare parts to entering the factory.
Supply and demand forecast
Forecast the future sales or some short time supply and demand according the current sales data, to help
arrange transport capacity beforehand.
Intelligent disassemling orders Dynamicly and intelligently deliver the orders.
Loading optimization Optimization for loading and unloading,including the optimizaton of alternative loading and unloading.
Dynamic scheduling Dynamicly ajust transport schedule according to the real transport state
Storage optimization Supply optimal warehousing level according to sales and supplying ability of the supplying chain.
AI platform AI calculation platform
22
Medium-term planning
function
Plan scheduling
function
Order
combination
planning
Row
scheduling
design
Predictive
warning function
Data
prediction
Abnormal
warning
Short-term planning
function
Scheduling
Daily
scheduli
ng plan
Emergency
intelligent
scheduling
Load/Unload
line
Loading
and
unloadin
g
Route
planning
Time
flow
control
Long-term planning function
Long-term site
planning
location
strategy
Supply
chain
strategy
Wareho
using
Configuration
transformation plan
Network
configura
tion
Short
barge
admissi
on
Convec
tion
cycle
Overall customization
Module
combination
System
customization
Further help finish Supply chain algorithm solution for the
manufacturing industry, retail industry, etc.
物件
装卸
销售
线路
规划原材料
排程
调度
仓储
协同
决策
演化
销售
预测预测
场景
识别
识别
Optimization
decision algorithm
prototype model
library
Deep learning
prototype model
library
Meta learning super heuristic algorithm
framework
• Mathematical model
of the problem
• Data and annotate
• Initialisation
• Evaluation system
• Others
static
customizedalgorithm
Intelligent learning and decision customized algorithm development
platform
Algorithm framework/Basic
function and class
Algorithm
engineer
Selfadaptedcustomizedalgorithm
Seek the
solution
Modeling
24
 Contact:
– Xinjie Zhang :
 E-mail:zhangxinjie@hotmail.com
 Tel:13311931669
 http://www.jzlog.cn:8000/
– https://istbi.fudan.edu.cn/lnen/
Thanks!

Smart logistics solution

  • 1.
    —Introduction to thesmart logistics project Shanghai Jiuzhang Intelligent Technology Co., Ltd. The Institute of Science and Technology for Brain-inspired Intelligence (ISTBI), Fudan Transport All Goods Easily
  • 2.
  • 3.
    3 Artificial intelligence Internet of things UnmannedBig data Logistics:Real world supply chain management  AI logistics target levels and constraint levels  AI logistics digitalization, platform based,automation  Project content and innovation points  Project target exhibit  Application cases in real world:How AI help logistics suppliers realise better social targets Project target:help realise intelligent logistics which meets real world needs
  • 4.
    4 Transportation strategy Warehousing strategy Networkstrategy Dispatching control Forcast group Decision making planning Cost Output Budgetary constraints Institutional constraints AI strategy  Massive variables  Multi-dimension  Multi-size  Dynamic  Multi-target  Strong constraint  Non-formulization  Non-Structured data Maximize service & minimize cost under constraint conditions for logistics optimization
  • 5.
    5 AI helps grapplewith the challenge of the future logistics management development Automation Optimization algorithm Machine Learning  Unmanned warehouse  Unmanned transportation  Overall industries supply channel integration  Real time and dynamic  Non-structured logistics big data mining  Forecast
  • 6.
    6 Project characteristics andinnovation points Full supply chain (B2B+B2C/road,rail,waterway) Dynamic and real time Non structural data graphical representation Brain-inspired intelligence new algorithm new model experiment platform Based on current situation, Look for future (Artificial + intelligence ) Artificial intelligence  Enhanced artificial intelligence Optimization/enhanced learning as the core
  • 7.
    7 Classic case:Some large-scaleautomobile logistics company project outline • Production line material distribution • After-sales logistics • Warehousing planning • Parts logistics Warehousing Capacity RouteScheduling Data Optimiz ation Study Forecast
  • 8.
    8 Classic case:Spare partslogistics transport optimization learning algorithm  Spare parts logistics transportation optimization needs overall scheme :Transportation route optimization,loading optimization,and queuing optimization.  Overall optimization algorithm supports: 100,000 orders per day,with mileage of 10 million KM.
  • 9.
    9 Classic case:constraint conditionsmulti-target combinational optimization distrib ution distrib ution distrib ution Loading diagram Route diagram Loadi ng Queuing(Flow control) Unloa ding
  • 10.
    10 Classic case :difficulties 1.Large number of variables 2. More restrictions 3. Nonlinear , unformulated 4. Large solution space : >1062 algorithm 1. Accuracy 2. Adaptive, learnable strategy updates 3. Intelligent algorithm and business adaptation coordination Business 13 Loading constraint 8 Transfer library constraint 12 Loading constraint 8 Resource constraint 8 Time window constraint 14 Multiple rounds of transport constraints constraints
  • 11.
    Innovation points:Solutions tooptimization algorithm + learning algorithm Intelligent Logistics Simulation Module Standardiz ation interface Overall intelligent optimization search algorithm Analog backout Inheritance algorithm Taboo search Parallel interfac e Machine learning module SVM Reinforceme nt learning 1. W. Lu et al Intelligent Solution System towards Parts Logistics Optimization ERUO2018 (Abstract, Contributed Talk) 2. Y. Huang, B. Chen, W. Lu et al. Intelligent Solution System towards Parts Logistics Optimization WCGO2019 (Regular Paper Accepted)
  • 12.
    12 Classic case: algorithmscombination of sucessfully researched and developed AI logistics ① SVM——Disassemble order strategy ② Greedy algorithm——Orders preprocessing ③ Community clustering——Initialization ④ Parallel analog backout algorithm ⑤ Self-adaption backout algorithm ⑥ Pruning searching algorithm with priority queue——Route planning ⑦ Dynamic planning——Loading + route planning ⑧ Intensive learning,Q-Learning——Loading + route planning ⑨ Tree search algorithm——Post-optimization+route combination Generality algorithm development Professionality algorithm research and development
  • 13.
    Classic case:the resultbetter than some world famous logistic management software Result of a famous business software Our result Load mileage 42235.5 39857.8 total mileage 84006.4 Dynamic loading rate 55% 52% Number of vehicles 118 155 Number of cars 315 Running time 30分钟 10分钟 number of orders 2959 2959 The system under algorithm has been at online operation.
  • 14.
  • 15.
  • 16.
    16 Business mode:Through technologyinnovation cooperation,realise economic and social benefits APP Platform service Vehicle & goods match Transport management AaaS Algorithm:providing customized algorithm for logistics enterprises Algorithm Scheduling algorithm,route optimization Site selection service ,Loading service Economic benefits : Reduce carbon emission Reduce source consumption Social benefits : Realize cost-reducing & increase benefits in whole society Increase economic operation efficiency Independent innovation: Basic algorithm theory innovation Basic software indepentent innovation SaaS Algorithm:providing cloud/personalization service mode
  • 17.
    17 1、Develop a platformfor optimization of vehicle and goods matching:Goods which can walk automaticly • Freight transportation platform like Didi or Uber • Intelligently matched order • Intelligently arranged route and vehicle dispatch • Route optimization • Increase the income of the freight carrier • Reduce no-load rate • Reduce carbon emission • Reduce transactional cost of vehicle and goods matching • Reduce loss and resource waste
  • 18.
    2、Develop SaaS System:Flexiblely embed existing logistics management system AaaS : jiuzhang VRP Input: Information of vehicles Orders info Distance matrix Output: Scheduling Optimized routine
  • 19.
    3、AaaS(Algorithms as aService) Service of consultation plus algorithm provided according to the clients’ business pain points. Rrequirement survey Interview research Data survey Establish a plan Modeling Adjust algorithm Conduct tests Select Range Test summarisation Promotion plan Training and extension Summarize and improve starting
  • 20.
    20 Future development andexpectation 算法开发 场景模拟 应用验证 “Strong” Artificial intelligence and brain- inspired intelligence Business→Mathe matics→ Business  Unmanned equipments  Transport method integration and optimization(highway、 railway、air transport、 waterway transport)  B2B and B2C integration  Coordination trans legal person and industries (Connect data,platform and equipments. )
  • 21.
    21 Future overall optimizationof logistic supply chain, supporting the full load of a vehicle, express transportation、spare parts, delivery,etc. Solution Description Route planning Route optimization for multi-vehicle,multi-goods and multi-sites.Include two categories:1,Without alternative loading and unloading. 2,With alternative loading and unloading(will load halfway) Load planning How to load the goods into the vehicle. 1,With the highest space use rate. 2,In the order of loading and unloading. Include:1,Loading schedule with the order of loading and unloading. 2, Loading schedule without the order of loading and unloading. 3,Consider the center of gravity for the boxes transported in UAV. Node location Choose intermediate supply point(or the front warehouse). In order that, 1,New network under logistic load is the most economical.(distance, cost) Entry and exit queuing management According to the vehicle entry time,unloading time,and with the help of the electronic fence,arrange the order and time of vehicle entry and exit. Reduce the traffice jam at the gate of factories. Three dimensional volume calculation With photo or surround video taking, 1,calculate the standard cuboidal object’s 3-dimensional volume. 2, Determine the irregular objects’ outsourcing line. Parts entering the factory According to the spare parts’ location,route schedule and manufacturing plan,decide the transportation plan and dynamic scheduling of the spare parts to entering the factory. Supply and demand forecast Forecast the future sales or some short time supply and demand according the current sales data, to help arrange transport capacity beforehand. Intelligent disassemling orders Dynamicly and intelligently deliver the orders. Loading optimization Optimization for loading and unloading,including the optimizaton of alternative loading and unloading. Dynamic scheduling Dynamicly ajust transport schedule according to the real transport state Storage optimization Supply optimal warehousing level according to sales and supplying ability of the supplying chain. AI platform AI calculation platform
  • 22.
    22 Medium-term planning function Plan scheduling function Order combination planning Row scheduling design Predictive warningfunction Data prediction Abnormal warning Short-term planning function Scheduling Daily scheduli ng plan Emergency intelligent scheduling Load/Unload line Loading and unloadin g Route planning Time flow control Long-term planning function Long-term site planning location strategy Supply chain strategy Wareho using Configuration transformation plan Network configura tion Short barge admissi on Convec tion cycle Overall customization Module combination System customization Further help finish Supply chain algorithm solution for the manufacturing industry, retail industry, etc.
  • 23.
    物件 装卸 销售 线路 规划原材料 排程 调度 仓储 协同 决策 演化 销售 预测预测 场景 识别 识别 Optimization decision algorithm prototype model library Deeplearning prototype model library Meta learning super heuristic algorithm framework • Mathematical model of the problem • Data and annotate • Initialisation • Evaluation system • Others static customizedalgorithm Intelligent learning and decision customized algorithm development platform Algorithm framework/Basic function and class Algorithm engineer Selfadaptedcustomizedalgorithm Seek the solution Modeling
  • 24.
    24  Contact: – XinjieZhang :  E-mail:zhangxinjie@hotmail.com  Tel:13311931669  http://www.jzlog.cn:8000/ – https://istbi.fudan.edu.cn/lnen/ Thanks!