Abstract:

According to The Large Truck Crash Causation Study in 2015

13% of crashes involving heavy commercial vehicles are occurring because of the driver's tiredness, 25%–30% of driving accidents are related to drowsiness, remaining percentage is due to different reasons like loose brakes and road slipperiness and other reasons.

Introduction

Fleet management includes many activities those lead to improving fleet efficiency, higher revenue and less expenses. Using data analysis to detect flaws and predict issues before happening is the core of this research. The following diagram shows a brief about the analysis phases, but not inclusive, so please keep going through remaining sections.

This paper will tackle main 4 points:

1- Driver behavior:

2- Vehicle Status.

3- Trip Advisory

4- Emergency management.

I’ll discuss some hypothesis and try to lead practical experiments to differentiate between correlated factors vs. caused impacts. Which will result in clear advices to the customer in many levels.

(I delayed data sources mentioning so readers who are interested in ideas but not in technical details won’t be bored.)

1- Driver behavior:

1.a- drivers ranking:

This analysis shouldn’t be only ranking drivers against each other, which will be unfair, it will rank driver behavior over different situations and time frames, to detect the best working environment for every driver. And to lead a scientific statistical experiment, I will include an much related data in the analysis like historical weather data, and vehicle history if possible.

Input to analysis:

1- Vehicle information.

2- Trip information (source, destination, shipment load, distance, time, …)

3- trip range (highway or inside city or highway and passing through cities)

4- road information.

5- Driver data (age, experience, history, main health issues reported at the time of trip)

6- Weather data is coming from published resources

Output of Analysis:

1- Rank of drivers per same weather information.

2- Rank of drivers per same trip circumstances.

3- Rank of drivers per similar vehicles and loads.

4- Rank of drivers per similar roads conditions.

5- Rank of drivers per similar driving times (day/night shift)

6- Rank of drivers per trip category (inside city/ highway)

7- List of drivers and best working conditions where his performance will be up to peak. (early bird vs night owl, highway vs inside city, short trip vs long trip, best region, best weather , best vehicle type)

8- Monitoring of each driver behavior and early detection of major changes. (health issue or rebellion)

9- Fraud detection.

Important notes to be considered in this analysis:

1- Inside main cities in Saudi Arabia, there are a lot of road maintenance and exchanges, and car accidents, this will impact trip time, this should be considered as outliers and removed from input to come to more precise and fair analysis.

1.b- Tiredness and drowsiness detection

there are many famous analytical models created to predict driver tiredness and drowsiness, some of them are adopted in major car manufacturers.

This will depend on data coming from wearable devices, but I not recommending it as it will face resistance from drivers. Or from driver mobile devices.

And if both are not feasible, we can depend on input data from previous analysis (driver behavior)

Output :

1- Alarming the driver about his tiredness.

2- Suggestions for driver to have a rest, and forcing him if his behavior exceeded a predefined threshold. Guiding him to nearest rest room or hotel.

3- Escalation if driver continue on his trip.

2-Vehicle Status and Utilization

this part value is underestimated although it’s very important, vehicle bad status is a major reason of car accidents, correct maintenance will save huge amounts of expenses (including destroyed cars and shipments, insurance, and even maintenance cost itself) and more important it saves souls.

Utilization is most of current customers are focusing on, although it is lacking a lot of features that can be provided to customer.

Input to analysis:

1- Vehicle information (model, type, mileage, …).

2- Vehicle maintenance history.

3- Trip information (source, destination, shipment load, distance, time, …)

4- trip range (highway or inside city or highway and passing through cities)

5- road information.

6- Vehicle feed (speed, tier pressure, fuel meter, load, ….)

7- Weather data is coming from published resources

Output of Analysis:

1- Vehicle Utilization report. (include over usage and under usage)

2- Prediction of next vehicle maintenance time. (periodic maintenance like oil change, brakes renewal,…)

3- Prediction of issues in vehicle, suggesting maintenance and even suggesting spare parts. To be bought and ready

(this will reduce wasted time of stopped car waiting for spare parts to be shipped from abroad)

4- Prediction of the need to exchange tiers (even if their mileage is not consumed yet) From harsh braking and tier pressure change rate. Due to problem in manufacturing or big load of trucks or hot weathers or bad roads.

5- Fuel mileage report: Get an estimate of the fuel usage by vehicle for a specified period of time and identify problems if consumption is far beyond the mean.

6- Maintenance reports: monitoring vehicle performance before and after maintenance and report the quality of maintenance done.

3-Trip Advisory:

this is my favorite part, as it depends on less input from the customer, required no additional devices and can generate a lot of valuable reports to increase revenue.

Input to analysis:

1- Trip information (source, destination, shipment load, distance, time, …)

2- trip range (highway or inside city or highway and passing through cities)

3- weather data

4- road data

5- places related to company (inventory, customer maintenance workshops, ..)

6- locations of gas stations, maintenance stations, emergency points, and rest houses or hotels.

Output of Analysis:

1- efficiency of the loading/unloading process in customer inventories (may customer needs more forks if loading/unloading process impacting car utilization)

2- efficiency of maintenance process (time taken to check, fix, exchange per vehicle model and part)

3- report about known obstacles, Customs or border points.

4- Prediction of random obstacles: due to traffic congestion, or sometimes when traffic officers forces trucks to part temporarily when there is a traffic jam, this is especially inside city trips, so drivers will be informed to avoid collisions.

5- suggest the routes to customers.

6- suggest routes taking maintenance workshops in consideration if the vehicle maintenance time is close, or vehicle behavior indicates the need for maintenance.

7- Suggesting routes taking gas station in consideration if fuel level is low.

8- Detection of gas station problems (no fuel for example) so alarming drivers to fuel earlier.

9- Detect Excessive stop or standstill times outside of an identified point

4-Emergency Management

it would be great if STC provided an emergency center which will do the following:

1- Receive any emergency calls from drivers (tank empty, fire, exploded tiers with no backup, driver health issue)

2- Warning cars on duty if there is any expected weather change, like (sand storm or heavy rains.)

3- Monitoring cars on road and detection of any deviation from normal road. It may be due to new path for the driver or kidnapping operation.

Requirements:

to complete our data set for analysis , we must include more data rather than fleet feed

1- There are some available weather data sets published over internet.

https://www.timeanddate.com/weather/saudi-arabia/riyadh/historic?month=7&year=2017

2- Some information should be collected from customer

Existing : car model, usage, maintenance ,
driver health history.
coming from additional sensors (tier pressure, gas tank meter, co1 and co2 from car exhaust, …..)

3- Some information should be collected

Gas station locations
emergency maintenance locations.

Hint:

We can conduct all of mentioned analyses in this paper internally in the company, actually, most of them are 1 man job to small team tasks. Major show stopper is compliance of customers with revealing their information or adding required devices and attachments to generate accurate data, and if they keep history of the required information.