Keywords

1 Introduction

Due to the more demanding quality requirements for manual transmissions imposed by the market in the last years, driver sensation during a gearshift has gained a very high importance during the design phase of the shifting systems once it is one of the most remarkable actions in terms of comfort perception in motor vehicles.

Actually, the first article found in the available literature about shifting analysis dates from 1949 and was published by M’Ewen [11] with the very first studies done in this front with the objective to understand the mechanics/behavior of the synchronization system in order to improve and ease the gearshift of war tanks in battle fields. This article is recognized as the first “theory of manual gearshifting”. After M’Ewen [11] just one article was found that considers a modeled drive which controls the velocity of the lever during the engagement phase. Kim et al. [2] detailed the dynamics of the synchronizers and gearshift system, applying a proportional, integral and derivative (PID) controller to represent the driver.

Another work performed by Hannemann [5] presented a statistical analysis based on objective results measured from a real situation which aimed to remove the inherent subjective portion of the real driver and his/her respective ergonomics/human factors. The author proposed a non-linear regression to estimate the subjective rating of the drivers for a given vehicle. However, although the author obtained good results, one of the proposed next steps would be to extend this analysis for other countries due to differences among countries around the globe.

Having said that, it is possible to verify in the available literature that no human interaction analysis with the gearshift system considering human factors and ergonomics (HFE) to define its relation to the physical values observed during the process of gearshifting (e.g. efforts, impulses, times, etc.) to support changes in the transmission hardware.

About gearshift quality, one can easily find several definitions, but definitely all lead to a not so clear definition and has been evolving in the last years mainly due to the more demanding market requirements [3, 5, 6].

Moreover, since the pioneer study performed by M’Ewen [11] until the ones presented by Szadkowski [16] and Szadkowski and McNerney [17], the main concern was in the analysis of the influence of the maximum effort demanded to shift gears, and nowadays with the addition of the shift comfort portion [3, 5, 6].

From the studies developed by Duque and Aquino Jr. [3], the model proposed based on HFE aspects plays an important role in the gearshift quality perception and, consequently, the subjective rating given by the driver.

Therefore, the objective of this work is to develop an in-vehicle test protocol, which applies surface electromyography (sEMG) techniques together with anthropometric aspects analysis of the test drivers.

2 Human-System Interface Applied in Manual Gearshifting

According to Czaja and Nair [2], a system can be defined as a set of organized and structured elements, possibly arranged in a hierarchic form, in order to perform tasks and to achieve its objectives. In addition to that, the authors mention that human and machine elements commonly compose a system set to achieve a common aim, with inputs and outputs, and limited by boundaries to define its separation to external and internal elements.

Applying this definition to the vehicle interface, the interaction between driver and vehicle is very wide and cannot be limited in studying only his/her contact with the gearshift lever inside the passenger compartment. However, take into consideration all interactions the drive may suffer while driving would demand a huge amount of time and more detailed and specific studies [6, 18, 19].

Said that, this research deals just with the controls reachability within passenger compartment, being focused on the gearshift lever. Exactly as cited by Robinette [15], reachability analysis is performed through a set of in-vehicle anthropometric studies, and must be the very first activity when designing the interior of a given vehicle [14]. In the available literature [12, 14], all reachability analyses done in a vehicle just guarantee that, statically, a population of drivers is able to reach and operate its controls with comfort and without facing any difficulty. However, these sort of analyses do not consider the loads that, as for example, the gearshift system transfers to driver’s hand during a gear change, similar to Parkinson and Reed [12] proposed.

Loads analyses are studied measuring the forces behavior in the lever knob during the entire gearshift, covering both selection and engagement phases [1, 3, 4, 6]. After running these measurements, the gearshift engineer decides if a change is necessary to improve quality, without considering any HFE aspect.

In this work, this analysis is replaced by measuring the muscular activity during the gearshift through sEMG signal acquisition. The details of the protocol applied and the equipment required are further detailed in the next sections.

3 sEMG Methods

According to Konrad [9], kinesiological sEMG can be defined as a discipline that studies the neuromuscular activation during functional movements, work situations and medical treatments. The motor unit is composed by a set of fibers, which are spatially positioned in a random way in different distances from the acquisition sEMG electrode, as shown in Fig. 1.

Fig. 1.
figure 1

(Source: Konrad [9], p. 9)

MUAP representation

Full size image

In case of several motor units being activated, the final signal, also known as motor unit action potential (MUAP), will be the sum of all resulting signals from each motor unit [8, 13, 20]. In addition, the MUAP’s of all detected motor units are observed as a symmetric distribution with negative and positive amplitudes with zero as average value.

Another point raised by Konrad [9] is that, although the human body is considered as a good electrical conductor, variations on conductivity are observed due to type of tissue, thickness, physiologic changes and temperature, as represented in Fig. 2.

Fig. 2.
figure 2

(Source: Konrad [9], p. 12)

MUAP differences representation

Full size image

These differences totally forbid the direct comparison among drivers, requiring a normalization of the sEMG signals based on the maximum voluntary contraction (MVC) [8–10].

The normalization of sEMG signals will be further discussed in Sect. 4 to define the post-processing methods used in this research.

4 Equipments, Test Protocol and sEMG Post-processing

For the sEMG signals acquisition were used the module ML880 PowerLab® 16/30 and the amplifier ML138 Octal Bio Amp®, both manufactured by ADInstruments with the configuration and post-treatment software LabChart® v7.2 (see example in Fig. 3). The equipment was set with a ±5 mV maximum amplitude with an acquisition rate of 1 kHz and filtered within the range of 20 to 500 Hz for the recorded channels.

Fig. 3.
figure 3

(Source: Author)

In-vehicle sEMG equipment installation

Full size image

Cables connect the acquisition module to the surface electrodes of the selected muscular group. With all cabling connections done, the driver sits into the test vehicle, adjusts the seat in a comfortable position, but in a straight posture to avoid any secondary movement during the gearshifts. After these set of adjustments, with the vehicle at rest and engine off, the driver performs some static shifts just for his/her familiarization with the equipment, being used also for the quality check of the sEMG signals. With the sEMG signals behaving accordingly and the adaptation phase concluded, prior to the test trials an acquisition of the MVC for the normalization of the sEMG signals is performed. The MVC acquisition is performed with the vehicle still at rest and the driver applying the maximum effort at the following gearshift lever positions:

  • 1st/2nd selection;

  • 5th/6th selection;

  • Engagement position of all gears, reverse included;

  • Reverse inhibitor activation only.

The time duration of each measurement was about 8 s in order to obtain a stabilized sEMG signal. After the MVC test conclusion, the driver starts the engine and drives the vehicle in a known circuit shifting gears following the vehicle speed profile seen in Table 1.

Table 1. Vehicle speed profile per shift.
Full size table

The sEMG signals are collected and the driver gives his/her subjective evaluation following the ATZ scale presented in Fig. 4, rating the quality perception for each shift.

Fig. 4.
figure 4

(Source: Hau [6], p. 40)

ATZ rating scale

Full size image

Due to the availability of the equipment and the drivers, the experiment had to be divided in two data sets, covering different variables. The first one, from now on called test A, was done with 8 drivers (7 female and 1 male) and was performed to evaluate the impacts on human subjective rating due to their differences in terms of anthropometric dimensions.

Regarding test B, just one driver was selected to perform the sEMG signals acquisition to correlate with his subjective rating about quality perception. Only in this test, even that no comparison would be made with other drivers, the MVC experiment was performed in order to validate the entire test protocol.

4.1 Anthropometric Measurements

The body dimensions chosen for this research are the height of the driver, and the length from the contact point of the heel with the ground to H-point and from the H-point to the right shoulder (see Fig. 5).

Fig. 5.
figure 5

(Source: Author adapted from InnerBody [7])

Body dimensions measured

Full size image

The required measurements are performed prior to the test, just after the briefing done to the driver about the entire experiment.

4.2 Muscles Selection and sEMG Acquisition Preparation

The surface electrodes were positioned in driver’s right side Pectoralis Major muscular group, as shown in Fig. 6, following SENIAM recommendations, with the reference electrode being located in the olecranon.

Fig. 6.
figure 6

(Source: Author adapted from InnerBody [7])

Pectoralis Major muscular group

Full size image

For the preparation of the interested area, to obtain a good sEMG measurement depends on the condition of driver’s skin and electrodes positioning [9].

Regarding skin condition, for evaluations with high dynamic loads it is mandatory to adopt rigid cleanness methods, like the removal of superficial dead cells and use of abrasive cleaning pastes, according to Konrad [9]. Once in this experiment only small/slow movements are performed, a simple alcohol cleaning could be done with hair removal, if needed, to guarantee full electrode adhesion in the area of the target muscle (Fig. 7).

Fig. 7.
figure 7

(Source: Author adapted from Innerbody)

Surface electrodes

Full size image

4.3 sEMG Data Post-processing

After the tests, two sets of sEMG data were obtained: one file for the MVC and another file for the regular driving condition. Both files originated by the LabChart® software had to be first converted into two MAT files to be further treated by using scripts created with Matlab® R2013b. As mentioned before, once raw sEMG signal has a mean value around zero, a rectification process had to be applied in all channels. After the rectification, due to the inherent random behavior of the raw sEMG amplitudes, the use of smoothing methods are required in order to obtain the mean trend of the signal. The method chosen was the root mean square (RMS), being recognized as a very good choice for smoothing sEMG signals [8, 9].

This process was performed for all shifts, covering both MVC and regular driving conditions. Moreover, after the RMS calculation for both datasets, it was possible to normalize the signals from regular driving trials.

5 Results

All statistical work in this study was performed using an unidimensional linear regression per selected variable calculated through the statistical tools from Microsoft Excel® 2010.

The way to check the correlation of the selected variable versus the given subjective rating is done by the \( \overline{R}^{2} \) (a.k.a. adjusted \( R^{2} \)), while the checks of the statistical relevance of the linear estimator \( \hat{\beta } \) is done by the calculation of p-value considering a 95 % confidence interval. Table 2 shows the statistical results from test A, which is possible to see that the correlation of the linear models based on anthropometric data were very poor, reaching a maximum value of about 3 %.

Table 2. Statistical results from test A
Full size table

Additionally to that, the p-values show that the linear estimators for all models do not have statistical relevance, which means that cannot be applied for estimating the subjective rating from bigger populations. About test B, Table 3 summarizes the statistical analysis.

Table 3. Statistical results from test B
Full size table

For test B, \( \overline{R}^{2} \) showed a better value, but still too low in terms of pure correlation of the linear model against the explained variable. Meanwhile, the calculated linear estimator showed a very low p-value, meaning that this model has relevance, statistically-wise.

6 Conclusions

Test A results showed that a linear model based only on anthropometric data of the driver does not correlate well to the subjective rating for shift quality perception. Moreover, the model does not present a good statistical behavior failing to generate a linear estimator within the 95 % confidence interval and cannot be used to predict the quality rating to given population.

On the other hand, test B results provided a linear estimator with statistical relevance, and could estimate a bigger population rating. However, the calculated \( R^{2} \) was very low with just about 13 % of correlation to the real rating.

It means that the model considering the sEMG signal of just Pectoralis Major group could not be used “alone”, requiring to be complemented by other variables to explain better the human rating.

Therefore, as a next step of this study, the authors recommend to identify other muscular groups that could be added and propose a multidimensional model instead of the simpler unidimensional approach. Another alternative linked to this one is to follow the non-linear format adopted by Hannemann [4] to define his shift comfort model, leading to a better correlation than the linear approach.

Additionally to that, other HFE-variables, like tactile-sense based, could be further analyzed and added to this new multidimensional model [3]. The results presented are being used in a research project that includes virtual reality interfaces and tele-existence [4] for monitoring data. In this case, the virtual reality resources are also used for simulation test environments.