Improving Efficiency in Transportation Networks using Python

Into The Logisticverse:

Presented by
Uzoma Nicholas Muoh

About Me

Your friendly neighborhood data nerd

$1.26T

Contribution by the Transportation industry to USA GDP as of 2022

$875.5B

Annual revenue of the U.S. Trucking Industry in 2021

Substantial Impact...

Statista. (2022)

Zippia. (2023)

72.5%

Market share of freight moved in the U.S by trucks.

10.9B tonnes

Total weight of freight transported by trucks in 2021

Trucking Industry

24.5%

Fuel consumption of commercial trucks in the US in 2021

American Trucking Associations. (n.d.)

American Trucking Associations. (n.d.)

U.S. Energy Information Administration. (n.d.)

Over-the-road Trucking

A type of trucking where drivers transport goods over long distances, such as interstate or international routes. 

At the đŸ«€ of the Problem

Carriers

These are the companies that pickup freight from a location and delivers it to the destination.

  • More customers and more orders

  • Fewer empty miles

  • Keep existing drivers and hire new ones

What Do They Care About?

i.e J.B. Hunt, Landstar

Shippers

These are companies that need their frieght moved.

  • Availability of trucks and drivers

  • Cheaper rates

What Do They Care About?

i.e P&G, Nestlé

Drivers

These are either owner-operators meaning they own or lease their own truck or they are employed by a carrier or broker.

  • Pay

  • Time

  • Being Safe

What Do They Care About?

The Carrier Network

Graph Analytics

A type of data analysis that focuses on studying the relationships between different entities which could be products, customers or even devices.

Graph Analytics

Undirected

Directed

Nodes

Edge

from io import StringIO
import pandas as pd
import networkx as nx

freight_activity = """
pickup,delivery,cost
A,B,13
A,C,14
A,D,10
A,E,10
A,G,16
B,C,12
C,A,14
E,F,30
E,G,23
G,A,16
"""

graph_df = pd.read_csv(StringIO(freight_activity))

graph = nx.from_pandas_edgelist(
    df=graph_df,
    source="pickup",
    target="delivery",
    edge_attr=["cost"],
    create_using=nx.DiGraph,
)

nx.draw(graph, with_labels=True)

Graph Analytics

 pip install networkx

Improving Connectivity

...
for node in graph.nodes:
    if graph.out_degree(node) == 0:
        print(f"Leaf Node: {node}")

Improving Connectivity

\alpha = \frac{e-v}{\frac{v(v-1)}{2} - (v-1) }
\beta = \frac{e}{v}

1.43

0.2

1.6

0.6

Collaborative Decision Making

Graphs for the Win đŸ€˜đŸŸ

Graphs

Analytics

Finding The Right Order

Shippers

Carriers

The Challenges

Shippers

Carrier

Factors, Components & Ratings

Components

Factors

Rating

Factors & Components

  • Financial
    • Order Throughput
    • Shipper Payment Turnaround
  • ​​​​​​Order Attribute
    • Equipment Type

Rating Mechanism

Is this a good order for us?

It is mostly good.

Fuzzy Logic

pip install scikit-fuzzy

Fuzzy Logic

import numpy as np
import skfuzzy as fuzz
from skfuzzy import control as ctrl

# Input Variables
order_input = ctrl.Antecedent(
    "order_throughput",
    np.arange(0, 3500, 150),
)
pay_input = ctrl.Antecedent(
    "pay_turnaround",
    np.arange(-365, 60, 1),
)
# Output Variable
fin_ouput = ctrl.Consequent(
    "financial",
    np.arange(0, 100, 10),
)
# Defining triangular membership functions
order_input["poor"] = fuzz.trimf(order_input.universe, [0, 0, 1750])
order_input["average"] = fuzz.trimf(order_input.universe, [1200, 1750, 2250])
order_input["good"] = fuzz.trimf(order_input.universe, [1750, 3500, 3500])

pay_input["poor"] = fuzz.trimf(pay_input.universe, [-365, -365, -120])
pay_input["average"] = fuzz.trimf(pay_input.universe, [-120, -30, -10])
pay_input["good"] = fuzz.trimf(pay_input.universe, [-30, 0, 60])

fin_ouput["poor"] = fuzz.trimf(fin_ouput.universe, [0, 0, 50])
fin_ouput["average"] = fuzz.trimf(fin_ouput.universe, [0, 50, 100])
fin_ouput["good"] = fuzz.trimf(fin_ouput.universe, [50, 100, 100])
# Create Rules
rule1 = ctrl.Rule(
    order_input["poor"] | pay_input["poor"],
    fin_ouput["poor"],
)
rule2 = ctrl.Rule(
    order_input["average"] | pay_input["average"],
    fin_ouput["average"],
)
rule3 = ctrl.Rule(
    order_input["good"] | pay_input["good"],
    fin_ouput["good"],
)
# Create Fuzzy Logic System
fin_ctrl = ctrl.ControlSystem([rule1, rule2, rule3])
financials = ctrl.ControlSystemSimulation(fin_ctrl)

financials.compute()

financials.input["order_throughput"] = 2600.94
financials.input["pay_turnaround"] = 10

print(financials.output["financial"]) # 76.67

fin_ouput.view(sim=financials) # good

đŸ‘đŸŸ good

Making Decisions Easier

Accept

Reject

Fuzzy Logic

Accept

Reject

Give Up

Matching Powers to Loads

Power

Loads

Drivers
&
Truck

?

Fleet Dispatcher

What Is a Good Match?

Combinatorial Assignment Problem

Tasks, Jobs, Facilities, Drivers etc 

Workers, Processors, Locations, Loads etc 

Cost Matrix

Cost Function

import numpy as np

def calculate_cost(power, load):
    cost = (
        w1 * hours_of_service_match(power, load)
        + w2 * route_distance_match(power, load)
        + w3 * driver_preference_match(power, load)
        + ...
        + wx * some_other_criteria(driver, load)
    )

    return cost
  
cost_matrix = np.empty((2, 2))

for i, p in enumerate(powers):
    for j, l in enumerate(loads):
        cost_matrix[i][j] = calculate_cost(p, l)

Mathematically...

C_{i,j} = \sum_{i=1}^{m} \sum_{j=1}^{n} \sum_{k=1}^{N} w_k * m_{k,i,j}

Minimizing The Cost Matrix

Google OR Tools

pip install ortools

Minimizing The Cost Matrix

# Instantiate a SimpleMinCostFlow solver.
smcf = min_cost_flow.SimpleMinCostFlow()

# Add each arc with their respective cost.
for i in range(len(start_nodes)):
    smcf.add_arc_with_capacity_and_unit_cost(
        start_nodes[i],
        end_nodes[i],
        capacities[i],
        costs[i],
    )

# Add node supplies.
for i in range(len(supplies)):
    smcf.set_node_supply(i, supplies[i])

# Find the minimum cost flow between 
# node 0 and node 5.
solve_status = smcf.solve()

import numpy as np
from itertools import product
from ortools.graph.python import min_cost_flow
# if ortools==9.3.10497
# from ortools.graph.pywrapgraph import SimpleMinCostFlow
NUM_POWERS = 2
NUM_LOADS = 2
TOTAL_ASSIGNMENTS = NUM_POWERS + NUM_LOADS

start_nodes = [0] * NUM_POWERS
end_nodes = list(range(1, NUM_POWERS + 1))
edge_cost = []

for i, j in product(range(NUM_POWERS), range(NUM_LOADS)):
    cost = cost_matrix[i][j]
    edge_cost.append(cost)
    start_nodes.append(i + 1)
    end_nodes.append(j + NUM_POWERS + 1)

costs = [0] * NUM_POWERS + edge_cost + [0] * NUM_POWERS

start_nodes += list(range(NUM_POWERS + 1, TOTAL_ASSIGNMENTS + 1))
end_nodes += [TOTAL_ASSIGNMENTS + 1] * NUM_LOADS

capacities = [1] * len(start_nodes)
supplies = (
    [NUM_POWERS]
    + [0] * TOTAL_ASSIGNMENTS
    + [-NUM_POWERS]
)

Et VoilĂ , We have Matches

Take aways  đŸ„Ą

  • There is a significant complexity in the problems the trucking industry deals with, things are not as simple.
  • Formulating the problem and getting the data to solve it is just half the battle.
  • Improving efficiency in transportation can look very different depending on the business needs.

Using Python's rich scientific ecosystem of libraries, we looked at:

  • Graph Analytics with networkx
  • Finding the right order with skfuzzy
  • Assignment Problem with google's ortools

Questions & Resources

Fosstodon: @nicksspirit

Twitter: @nicksspirit

Code Example:

Fuzzy Logic

Assignment Problem

Graph Analytics

Transportation Network

Into the Logisticverse: Improving Efficiency in Transportation Networks using Python

By Nick Muoh

Made with Slides.com

Into the Logisticverse: Improving Efficiency in Transportation Networks using Python

When we think about what Python is for, we often think of things like analytics, machine learning, and web apps but python is a workhorse that plays a tremendous and often invisible role in our day-to-day lives, from medicine to finance, and even the transportation of goods from manufacturers to the shelves of our neighborhood stores. Transportation networks are highly dynamic, goods are always moving from point A to point B and money every minute is being gained or lost. Improving efficiency in a transportation network is critical to the survival of a business that provides transportation and distribution services as well as ensuring timely delivery of goods to customers. This talk examines 3 real-world examples of how Python is used to improve the efficiency of transportation networks, particularly we will explore: * Finding the optimal match between a driver and a load at the lowest possible cost using google's `ortools`. * Generating recommendations for macro level optimizations to a transportation network using `networkX`. * Helping the descision making process by answering the question "Should I accept this work?" using `skfuzzy`. Key Takeaways: * Graph analytics and data science concepts that facilitate getting goods from manufacturers to stores more efficiently and at a lower cost to businesses. * An appreciation of the complexity of the logistics industry and the

  • 68