Most fleets across the globe provide performance-based incentives to their drivers. It is believed to be a very effective method of encouraging chauffeurs to drive responsibly, while the company earns increased profit on account of reduced maintenance as well as reduced fuel consumption. This ranking is generally based on fuel mileage attained by the driver’s vehicle during the assessment period. However, this is a very crude method of rewarding drivers, since the mileage of a vehicle varies a lot with the age of the vehicle as well as traffic conditions.
In this study, we propose a new method to rate the performance of drivers. We evaluate the drive rather than the driver and provide a rating for the complete trip. This rating is based on four different factors – vehicle idling, engine/ vehicle speed, acceleration profile and deceleration profile. The criteria for calculating the ranking keep getting updated as the vehicle covers more miles, thus removing discrepancies due to vehicle condition. We conclude that the drive rankings attained through this method correlate extremely well with the fluctuations in mileage across vehicles, drivers as well as routes.
Fuel cost and maintenance of vehicles are the two biggest variable expenses for any fleet company. People have come up with different methods to reduce these costs; most of these methods are focused on proper training of drivers. Different incentives are provided to the chauffeurs for driving responsibly, with their performances judged mostly on the basis of their vehicle’s mileage.
However, vehicle mileage alone does not serve as a good enough indicator of how the vehicle is being driven. Road congestion and driving profile of the chauffeur should also be taken into consideration. For example, we look at two speed profiles of two different drivers on the same route at different times of the day, (1) & (2).
As can be seen, drive profile 1 is much smoother, whereas drive profile 2 has too many steep accelerations and decelerations. Looking at it, one can clearly observe that the first driver is a much better handler of the vehicle. Still, the mileage numbers for both the drives are similar (within 1 % of each other). This is due to the fact that the drive 1 has long stretches of vehicle idling due to higher traffic congestions. Now, this is something out of control of the driver, but unfortunately might still cause a reduction in his performance appraisal, if the fleet manager only looks at mileage numbers.
A NOVEL WAY
In this paper, we introduce a novel way of rating a drive, which depends on a variety of factors, including traffic conditions, acceleration and deceleration profiles of the chauffeur. We then describe unique ways of numerically calculating each of these factors, using only vehicle speed and inculcating these factors into a tangible score.
We validate this model by using test data from vehicles being tested by Altigreen. One simplistic way to achieve this is by comparing driver ratings across a series of drives with the mileage numbers achieved by those drivers. We expect a bad performance in each of these factors to bring down the mileage number of that particular vehicle during the course of testing and provide the fleet manager with the exact reason for the change in fuel consumption patterns.
As per our performance algorithm, each drive is given a rating out of 10 points. These 10 points are further split up into four different factors with different weights –
:: Mean engine/vehicle speed – maximum of 2 points
:: Vehicle Idling Time – maximum of 2 points
:: Vehicle Acceleration Profile – maximum of 3 points
:: Vehicle Deceleration Profile – maximum of 3 points
Whereas the first two factors depend to a large extent on road traffic and congestion conditions, the third and fourth factor depend completely on the driver behaviour; hence, a greater weightage for these two factors. The actual numbers are calculated using the following steps –
Rating for Mean Engine/ Vehicle Speed
Ideally, the engine speed should be used for this analysis. However, on many occasions, fleet managers only record the vehicle driving speed. Hence, we have kept this particular factor flexible; it can be calculated using either engine speed (RPM) or vehicle driving speed. The mean driving speed of the drive is calculated using the following formula –
The mean engine speed (eRPM) can be calculated by simply taking a mean of all recorded values. For our analysis, we have used the following ranking criteria for this particular factor –
where, Si are the different speed limits and RSi are the rating points given for the factor. The number of conditions in the logic above depends on the granularity of the vehicle data being recorded, and the sensitivity one wants to build in the results.
As per our experience of analysing three years of driving data minutely, ICE engines perform optimally within a certain rpm band. For diesel engines of the capacity of 1,000-1,500 cc, this band is generally in the engine rpm range of 1,900-2,100. Similarly, if one is using vehicle speed as the criteria, the vehicle speed band of 40-60 km/h can be assigned maximum rating points for vehicles of the above-mentioned size. The set points can vary, depending on a country’s expected drive profile. The mean vehicle speed is then updated at the end of every drive and kept in record for later comparison –
where, D1 is the previous distance of the vehicle, D2 is the distance for the present drive, T1 is the previous total time of the vehicle and T2 is the present drive time of the vehicle.
Rating For Vehicle Idling
This particular factor is calculated using the percentage time in which the vehicle is kept in idle condition (engine running & vehicle speed <1 km/h). The final rating is then computed out of a maximum of 2 points based on the following algorithm –
where Ti are the different thresholds for idling time percentage and RIi are the rating values associated with this parameter. Again, the number of loops can be varied based on the requirement and granularity of data being collected.
Keeping the vehicle in idle condition is always detrimental to the health of the vehicle. At the same time, depending on traffic conditions, it is impossible to have negligible idling times, especially on Indian roads. This negligible idling time limit varies from city to city, depending on traffic congestion.
Vehicle Speed Bands
For calculating the acceleration and deceleration-related factors, first the entire drive is split into multiple speed ranges. These depend on the country/ area in which the vehicle is being driven. For our analyses we have chosen the speed ranges as 0-20 km/h, 20-40 km/h and 40-70 km/h.
These ranges are further split into different speed bands of 5 km/h each. For instance, the 0-20 km/h speed range is split into four bands – 0-5, 5-10, 10-15 and 15-20. The reason behind this split is that the acceleration and deceleration requirements are very different for different speeds. Mostly, we require higher acceleration at lower speeds, whereas driving with such high accelerations at higher speeds would be considered as rash driving.
Rating For Acceleration And Deceleration Profile
For calculating the last two factors, we will need a few initial parameter values. These values are taken from our historical data for a new vehicle and then updated as the vehicle keeps on clocking more distance. These parameters include mean driving speed; mean acceleration for each speed band; mean deceleration for each speed band; standard deviation of acceleration for each speed band and standard deviation of deceleration for each speed band.
As per our experience, we have observed that for similar traffic conditions, these values remain very close for a particular class of vehicle, like diesel sedans, petrol hatchbacks, etc. Since these values are updated at the end of every drive, our logic ensures that this impact of historical data keeps getting diluted with more drives for a particular vehicle.
For each speed band of 5 km/h, high and medium violation thresholds are calculated using the following set of formulae –
where, Ma is the mean acceleration for that particular speed band, σ is the standard deviation of all acceleration points in that speed band, Md is the mean deceleration for that particular speed band, and σd is the standard deviation of all deceleration points in that speed band.
Once these thresholds are calculated, a count for the number of high violations (acceleration/deceleration > high violation threshold) and medium violations (medium violation threshold < acceleration/deceleration < high violation threshold) is made for each speed band of 5 km/h. The drive is then split into the three speed ranges: low speed range (0-20 km/h), medium speed range (20-40 km/h) and high-speed range (> 40 km/h). The total violation counts in these speed ranges can be calculated as follows.
where, TVCi = total violation count for a speed range, and VCb = violation count in a sub-speed band, b, within a speed range, i. These total violation counts are then converted into percentage by dividing the total number to acceleration points within the speed range. The percentage violation threshold can be set according to the type of vehicle and will remain constant (can be updated if required, based on the initial drives).
We generally have kept the thresholds at about 2 % for stronger violations and 5 % for milder violations. These thresholds are calculated using the large amount of driving data we have been analysing over the past three years. Based on the calculation of violation percentage in the above step, the driver ratings for acceleration within one speed range is decided as:
where, RAi are the rating points associated with vehicle acceleration profile, HiAccV is high acceleration violation percentage in one speed range (low speed, medium speed or high speed), MildAccV is mild acceleration violation percentage in one speed range, Acc_HL2 is high violation per cent higher threshold in a speed range, Acc_HL1 is high violation per cent lower threshold in a speed range, Acc_ML2 is mild violation per cent higher threshold in a speed range, Acc_ML1 is mild violation per cent lower threshold in a speed range. Similarly, the ratings for deceleration can be decided.
The mean and standard deviation values will need to be updated after every drive. This ensures that our rating reflects the behaviour of a particular vehicle and a driver is always compared to his past driving behaviour. The new values can be calculated as follows –
where, D1 is the previous distance of the vehicle, D2 is the distance for the present drive, T1 is the previous total time of the vehicle, T2 is the present drive time of the vehicle,Acc1 is the previous mean acceleration/ deceleration value for a particular vehicle speed band, Acc2 is the mean acceleration/ deceleration for present drive for the same speed band, S1 is the previous standard deviation of acceleration/ deceleration for a particular speed band, S2 is the standard deviation of acceleration/ deceleration of present drive for a particular speed band, AT1 is the previous total acceleration/ deceleration duration for a particular speed band and AT2 is the acceleration/ deceleration duration for a particular speed band for present drive.
TEST RESULTS FROM ON-ROAD DATA
In order to validate the usefulness of our drive performance algorithms, we gathered speed data from one of our test vehicles for a few days with two different drivers. From our past experience, we had observed the Driver A to be a much better driver as compared to Driver B, who drives a little harshly. We also ensured that the vehicle fuel tank was full at the start of every day. We then used our algorithms to assign a drive rating to each of these drives to check if there was a correlation between our rating and the mileage the vehicle attained every day. The results obtained are presented in (3) and (4).
From the tests, we observed the following:
:: The drive rating correlated well with the vehicle mileage; generally, we see a spike in mileage whenever a higher rating was obtained.
:: The mileage correlated the most with the driving speed of the vehicle (5). It can be seen that the drives with low speed rating have a lower mileage. For drives with the same speed rating, the mileage varies depending on other drive performance factors.
:: Acceleration and Deceleration profiles play a significant role in determining vehicle performance as compared to vehicle idling behaviour.
:: The numbers for driver B were pretty bad as compared to driver A. As expected, his acceleration and deceleration rating points were also much lower than driver A.
:: Using our drive rating, we were able to pinpoint the reason behind differences in mileage between drives.
:: Even though there were some drives in which the mileage for driver A were same as driver B, we were able to still see the harshness of the drives of driver B.
The purpose of this study was to provide fleet managers with a comprehensive method to judge the performance of a vehicle, on a drive-by-drive basis. Whereas the idling and mean driving speed-related ratings provide a good indication about the route congestion, the acceleration/deceleration-related ratings offer an excellent feedback on the driving behaviour of the chauffeur. Using the first two factors, a fleet manager should look at avoiding routes with higher congestion (low drive speeds and high idling times). A driver performing badly on the last two factors should be trained accordingly; ignoring such behaviour might result in an accident sooner or later, leading to high maintenance costs. The sharp deceleration profiles will also lead to brake shoes wearing out quickly, again adding to the maintenance budget of the fleet.
Again, as shown above there are many occasions when two drivers with different driving habits cover the same route and attain similar mileage numbers, but their drive profiles are hugely different. It is almost impossible to detect such scenarios with currently used mileage-based methods. Our methodology helps to weed out such instances and warn fleet owners, help prevent accidents, reduce wear and tear on consumables and keep customers satisfied.
The accumulated rating correlated very well with the vehicle mileage, thereby helping fleet managers understand the reasons for fluctuations in vehicle fuel consumption.
 “Drive-style emissions testing on the latest two Honda hybrid technologies” by Adriano Alessandrini, Fabio Orecchini, Fernando Ortenzi & Federico Villatico Campbell
 “The Effects of Driving Style and Vehicle Performance on the Real-World Fuel Consumption of U.S. Light-Duty Vehicles” by Irene Michelle Berry
VISHESH MEHRA is Product Manager at Altigreen Propulsion Labs in Bengaluru (India).
AKSHAY JAGTAP is Engineer at Altigreen Propulsion Labs in Bengaluru (India).