A Review on the Development of a Test Method for Motorcycle Autonomous Emergency Braking Systems

Type of the Paper: Conference Paper

A Review on the Development of a Test Method for Motorcycle Autonomous Emergency Braking Systems

Nora L. Merkel^1,*

¹ Würzburger Institut für Verkehrswissenschaften (WIVW GmbH), Germany; merkel@wivw.de, https://orcid.org/0000-0002-4865-368X

*corresponding author

Name of Editor: Jason Moore

Submitted: 15/09/2023

Accepted: 05/10/2023

Published: 06/10/2023

Citation: Merkel, N. (2023). A Review on the Development of a Test Method for Motorcycle Autonomous Emergency Braking Systems. The Evolving Scholar - BMD 2023, 5th Edition. This work is licensed under a Creative Commons Attribution License (CC-BY).

Abstract:

In the passenger car and truck sector, assistance systems that intervene in emergency situations and thus help to improve vehicle safety have already been successfully used for many years. Although motorcyclists are subject to a high risk of suffering severe or fatal injuries in road traffic, systems that actively intervene in emergency situations are not yet available in the motorcycle sector. One reason for this is that passenger car systems cannot easily be adapted due to the motorcycle specific single-track vehicle dynamics. There are characteristic challenges that set limits to the possible application of actively intervening assistance systems. Exceeding these limits when applying an assistance system on a motorcycle could result in the occurrence of new critical situations that are no longer controllable for the rider. Still, previous research concludes that assistance systems for motorcycles have the potential to increase riding safety and identifies autonomous emergency braking systems for motorcycles (MAEB) as one of the most promising technologies (Savino et al., 2013).

One major challenge in MAEB studies is the conflict of goals between the aim to optimize the effectiveness of MAEB by identifying maximum possible decelerations that can be applied in a safe way and the wish to evaluate ‘natural’ rider reactions to an autonomous braking intervention. For the latter, riders should not anticipate the autonomous deceleration in order to achieve unbiased results. Obviously, it is ethically unacceptable to determine feasible deceleration limits with unprepared study participants. Approaching these limits carries the risk of provoking critical situations. During the research described in the paper at hand, a multi-phase approach was developed, in order to overcome the trade-off between achieving maximum effectiveness of braking interventions by identifying maximum feasible decelerations on the one hand and on the other hand evaluating unbiased reactions of unprepared riders. While other research groups focus on urban riding scenarios at velocities up to 50 km/h in their MAEB research (e.g., Lucci et al., 2021), the investigations described here concentrate on higher velocities as they occur in rural scenarios.

Throughout the research described in this paper, the developed investigation method was exemplarily applied to a prototype MAEB system. The paper provides an overview of the major results of all three phases of MAEB assessment. The method proves to be appropriate and delivers promising results regarding the applicability of autonomous emergency braking systems for motorcycles in the evaluated scenarios. The reproducibility of the measured rider reactions creates confidence that the corresponding effects are predictable, which means that the rider behavior does not represent a completely incalculable safety-critical factor for the application of MAEB. The successful application of the method leads to the conclusion that it can serve as a basis for the release of systems that intervene in the longitudinal dynamics. It gives manufacturers and system suppliers the opportunity to systematically prove that their systems are controllable for end users and can be applied without causing additional risks. Thus, the method can contribute to the future use of safety-enhancing assistance systems for motorcycles.

Keywords: Motorcycle Safety, Autonomous Emergency Braking, Aeb, Maeb, Motorcycle, Rider Behavior, Rider State

Introduction

The objective of this paper is to summarize the research on MAEB that was conducted at the Institute of Automotive Engineering (FZD) at TU Darmstadt throughout the past years. It describes the development of a multi-phase investigation method that allows to identify appropriate autonomous braking maneuvers that offer maximum effectiveness while not causing unreasonable additional risk in an already critical situation. The method consists of three levels. Each level is represented by a study concept with a certain objective. These individual levels were developed, applied and evaluated in various research projects. As the single investigations focus on quite different aspects, they cannot stand for themselves when aiming for an evaluation of the applicability of a system. Only combined within the method described in the following, the investigations provide a comprehensive assessment of a particular MAEB.

An in-depth discourse of the contents of this paper and more details on the development and evaluation of the method can be found in the doctoral thesis ‘Investigations on the Applicability of Autonomous Emergency Braking Systems for Motorcycles’ (Merkel, 2020).

Initial Situation¹

A lot of important findings have already been generated from previous research in the field of MAEB. E.g., the potential of MAEB to increase safety for motorcyclists has been analyzed and proven based on a detailed evaluation of accident databases and simulations (Savino et al., 2013). In laboratory tests it was shown that riders are not significantly destabilized in their position on the vehicle by an unexpected deceleration (up to 3.5 m/s²) and tolerate the maneuvers (Symeonidis et al., 2012). However, these and other findings were gained in more synthetic experiments, e.g., a laboratory test on sled construction (Symeonidis et al., 2012), deliberately riding towards a stationary obstacle (Savino et al., 2012), or automatic braking during free riding with no obstacle with riders informed about the purpose of the test (Savino et al., 2016). It remains to prove whether the results still apply under realistic conditions, i.e., if riders do not expect an emergency braking situation or an intervention.

Still, in terms of maximizing the effectiveness of MAEB, the finding that riders can control and accept rather low decelerations is not sufficient. It is essential to determine the limit to which autonomous emergency braking maneuvers are still controllable for motorcyclists in order to fully exploit the potential of MAEB to increase safety. Moreover, it must not only be investigated whether corresponding interventions are controllable for riders, but also if they are still accepted. Without the acceptance of the riders, the implementation of MAEB in the market cannot be realized successfully.

In order to obtain reasonable results concerning the feasibility and acceptance of MAEB, it is not sufficient to confront only prepared riders with autonomous braking maneuvers, as it is done in most of the known research. It is assumed that riders who expect the intervention (even if they do not know exactly when it will happen) will not react in the same way as those who experience the situation without any expectation. Thus, for investigating how motorcyclists handle an autonomous braking intervention and how they succeed to stabilize the vehicle, only test persons who are not informed about the aim of the experiment beforehand can be considered.

Identifying maximum controllable decelerations to achieve the greatest possible effectiveness of the intervention contradicts this desire to investigate natural rider reactions. The identification of controllability limits requires to increase interventions to critical ranges and thereby bears the risk of loss of control, i.e., destabilization of the vehicle, possibly even falls, and is consequently not justifiable from an ethical point of view.

It is therefore necessary to develop a suitable procedure that decouples the goals of ‘identification of controllability limits’ and ‘investigation of natural rider reactions’. This enables both questions to be addressed without subjecting participants to an unjustifiable risk of injury.

Test Method Development

The test method development is based on the assumption that for the application of an autonomous braking intervention at maximum deceleration, the rider needs to be in a so-called ready-for-braking state. This means that there is a certain amount of body tension and both hands are at the handlebar in order to support initial forces and the rider is aware of the braking situation. In a situation that requires an autonomous braking intervention, these requirements are usually not fulfilled. I.e., the rider must be motivated to change from an arbitrary state to the ready-for-braking state by building up body tension, bringing the hands to the handlebar and developing situation awareness. In the following, this phase will be referred to as ‘transition’. The described process is illustrated in Figure 1.

Figure 1. Phases of the rider state prior to a braking maneuver (Merkel, 2022).

Strategies to bring the rider to the ready-for-braking state before applying maximum automatic deceleration could for instance be acoustic, visual or haptic warnings. The approach chosen in the research described here is to apply a partial deceleration to initiate the transition phase. This is based on the assumption that a low-level deceleration will intuitively lead to situation awareness and causes body tension and a movement towards the handlebar automatically. The approach unites two advantages: On the one hand, there is no necessity to interpret a warning first, the transition is initiated automatically and on the other hand, velocity can already be reduced before the rider is ready for maximum deceleration and thus the effectiveness of the MAEB can be increased.

Consequently, the aim of the research is to find out how an automatic deceleration must be designed to achieve an effective initiation of the transition and which levels of deceleration are already applicable before the rider reaches the ready-for-braking state without causing a loss of control.

The method starts with a determination of controllability limits during the first phase. As stated before, an approximation to controllability limits must not be performed with naïve participants. In this stage of the investigations, the test persons are expert riders. These experts are experienced professional riding instructors and trainers. They are assumed to be particularly suitable to assessing the skills of unexperienced riders.

To determine the most suitable design, three different deceleration profiles as illustrated in Figure 2 were investigated during the studies: a constant low-level deceleration (‘block profile’), a slow buildup of deceleration (‘ramp profile’) and a short braking ‘impulse’. The experts evaluate increasing levels of the autonomous decelerations concerning their feasibility for unprofessional motorcyclists. Starting at a low level (3 m/s² for block and impulse, 3 m/s³ for ramp), decelerations are increased for each intervention, until the experts assess the maneuver as no longer controllable for non-expert riders. This is repeated at various initial velocities, in order to identify possible velocity dependencies of the feasible deceleration level and for all three deceleration profiles.

Figure 2. Braking profiles investigated during the expert study.

The parameters for controllable decelerations identified during the expert study build the basis for the second step of the test method. In this phase, for the first time, unprepared participants experience autonomous braking interventions in a realistic emergency braking scenario (suddenly decelerating preceding vehicle). Based on reaction times, velocity reduction (compared to baseline experiments with no autonomous braking intervention) and user acceptance, the parameterized deceleration profiles are evaluated concerning their potential to increase safety for motorcyclists during emergency braking situations. This leads to the identification of the most appropriate intervention strategy.

In the third phase, which is again conducted as participant study with unprofessional riders, the method intends to analyze the reactions of riders to unexpected braking interventions in a more detailed way. It concentrates on the profile that was identified as most appropriate during the preceding study. The focus of this third study part is on the relative movements between the riders and the motorcycle. This includes, e.g., upper body and head displacement as well as support forces on the handlebar. A main focus is on the identification of characteristic behavior and timing. This helps to gain further knowledge about the requirements for the design of autonomous emergency braking interventions that result from characteristic rider behavior.

The three phases of the investigation method are summarized in Figure 3.

Figure 3. Three phases of the investigation method (Merkel, 2022).

Application of the Method and Results

Throughout the research described in this paper, the developed investigation method was exemplarily applied to a prototype MAEB system. All three study concepts described in the previous section were carried out on the August-Euler-Airfield in Griesheim near Darmstadt throughout various projects. The main results are summarized in the following.

Expert study – parameterization of deceleration profiles

First of all, the expert study was performed with five riding trainers and instructors. While riding at a determined velocity on a straight and flat section of the test track on a measurement motorcycle, they were knowingly decelerated via remote control. The automatic braking interventions were triggered via remote control by the experiment supervisor. The experts evaluated the controllability of the three deceleration profiles at increasing levels. The determined parameters build the basis for the following participant studies. The feasible decelerations or deceleration gradients were mainly assessed for initiating the deceleration at 70 km/h, as this is the intended velocity for the participant studies (representing riding in rural scenarios). As it cannot be ruled out that the initial velocity influences the controllability of deceleration interventions, the determined limits only apply for this explicit velocity. First investigations on the velocity influence can be found in Merkel et al. (2018).

The identified controllability levels for unprepared non-expert riders at 70 km/h are shown in Table 1.

Table 1. Results of the expert study (Merkel et al., 2018).

Braking profiles	Varied parameters		Determined controllability limit
Block	Level of deceleration		5 m/s2
Ramp	Gradient		9.1 m/s3
Impulse	Level of deceleration		4.7 m/s2

First participant study – identification of most appropriate deceleration profile

With the knowledge on controllable deceleration (gradient) levels for average riders, the first participant study was performed. The aim of this study was to examine to what extent the different deceleration profiles are suitable to assist riders in emergency braking situations and which increase of safety this can offer.

During the experiments, the riders followed a dummy target on the measurement motorcycle. At an appropriate instance (correct distance between the vehicles, correct velocity, enough straight track left), the deceleration was triggered by the experiment supervisor. Besides maneuvers with the three autonomous deceleration profiles, reference experiments without any intervention were performed in order to gain a ground truth by measuring unassisted reaction times. The study was carried out with 18 participants. With the aim of receiving unbiased assessments and to avoid habituation effects, only two braking maneuvers were performed per participant. After elimination of the invalid runs, 19 braking maneuvers can be evaluated (5x block, 5x ramp, 5x impulse, 4x reference). (Merkel et al., 2019)

First of all, the reaction times to the autonomous deceleration were evaluated. During a braking maneuver, the deceleration of the vehicle causes a forward displacement of the rider’s upper body relatively to the vehicle. Only when the rider adapts to the changing vehicle state, the deceleration can also be found at the upper body. This causes a certain time lag between the vehicle deceleration and the deceleration of the rider body. This time lag represents the transition time that riders need to reach the ready-for-braking state. The upper body deceleration is measured by an acceleration sensor that is mounted to the riders’ back. The Additional measurements that were taken into account in order to evaluate the transition were the rising supporting force on the handlebar and reactions in terms of applying the brake or clutch levers. These measures come directly from the measurement motorcycle. The transition times can be found in Table 2.

To evaluate the safety potential of the MAEB to increase safety for motorcyclists, the possible gain of velocity reduction during the reference reaction time was determined. The first pillar of this evaluation is the measured velocity reduction during the transition phase. The second pillar is based on a theoretical consideration: It is assumed that after the transition phase, it is possible to increase the autonomous deceleration to a maximum level. This means to determine the potential velocity reduction Δv_Red within the reference time T_Ref for each braking profile, it is assumed that after the transition period T_Trans, the rest of the reference time span (1.65 s) is used to decelerate at D_max = 7 m/s². The calculation is exemplarily shown for the block profile in (Equation 1). Within the transition time of 0.57 s, the velocity is reduced by 1.48 m/s (mean reduction determined during experiments). The rest of 1.08 s within the reference phase are used to decelerate at 7 m/s². This results in a total velocity reduction of 9.04 m/s. The results of this analysis are also shown in Table 2.

\[\mathrm{\Delta}v_{Red,Block} = \mathrm{\Delta}v_{Trans,Block} + \left( T_{Ref} - T_{Trans,\ Block} \right) \bullet D_{\max}\] \[\ \ \ \ \ \ \ \ \ \ \ \ \ \ = 1.48\ \frac{\text{m}}{\text{s}} + \left( 1.65\ \text{s} - 0.57\ \text{s} \right) \bullet 7\frac{\text{m}}{\text{s}^{\text{2}}}\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \] \[\ \ \ \ \ \ \ \ \ \ \ \ \ \ = 9.04\ \frac{\text{m}}{\text{s}}\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \]

(1)

Table 2. Potential velocity reduction of the deceleration profiles (Merkel et al., 2019).

Profile	Mean transition time TTrans in s	Mean velocity reduction within transition period ΔvTrans in m/s	Potential velocity reduction within 1.65 s ΔvRed in m/s
Block	0.57	1.48	9.04
Ramp	1.04	1.69	5.96
Impulse	1.37	0.77	2.73
Reference	1.65	0.57	0.57

The results show that the block profile is the most appropriate strategy in order to apply a partial deceleration before riders reach the ready-for-braking state. With an average of 0.57 s, it generates the fastest transition and enables a comparatively large velocity reduction already during the transition phase due to the fast brake pressure build-up.

In addition to the analysis of the measurement data, the participants were interviewed concerning their subjective evaluation of the braking interventions. The evaluation was based the controllability scale according to Neukum et al. (2008). Within this scale ratings are first done between ‘noticeable’, disturbing’, ‘dangerous’ and ‘not controllable’. Within these categories the interventions are then classified on a finer scale (see Figure 4).

Figure 4 shows the distributions of the ratings given for the different maneuvers. As expected, the reference scenarios are rated least critical. These maneuvers are mostly assigned to the lower end of the scale in the harmless range. All but one of the block braking interventions are also rated as harmless. However, the average of 2.8 is slightly higher than for the reference braking (2.67).

On average, the ramp profile is also rated as harmless (3.4). However, the spread is larger, with several ratings in the ‘disturbing’ range. The impulse profile is rated most critically. The average rating is in the lower ‘disturbing’ range (4). But here, too, there is a scattering of ratings across the ‘noticeable’ and ‘disturbing’ ranges.

Figure 4. User acceptance of deceleration profiles (Merkel et al., 2022).

Consequently, the block profile not only performs best in terms of transition time and velocity reduction (objective criteria), but also in the subjective assessment of controllability. For this reason, the second participant study concentrates on this profile.

Second participant study – analysis of characteristic rider reactions²

After the controllability limits for unintended decelerations and the most appropriate braking profile have been identified, the third study part (second participant study) has the aim to achieve a better understanding on how riders adapt to the changing vehicle state during a braking maneuver and thus identify characteristic behavior. In investigations on the identification of appropriate measures for such evaluations, the relative movements between the rider’s upper body and the motorcycle were found to be suitable to describe the rider reactions to sudden unexpected changes of the vehicle state. Within the second participant study, these measures are used to analyze the reproducibility of the rider behavior during braking maneuvers, including a comparison of automatic braking interventions vs. maneuvers in which the riders had to apply the brakes themselves (Merkel & Winner, 2020). The upper body movements are measured by three wireless inertial measurement units that are attached to the rider’s back and helmet as it is shown in Figure 5.

Figure 5. Sensor positions.

When the rider experiences an automatic braking intervention, due to inertial forces, the upper body moves forward. As the rider is connected to the motorcycle at the seat and the forward movement is limited by the fuel tank, this results in a pitch movement (see Figure 6) that can be measured at different points of the upper body and also at the head of the rider. We assume that the point at which the rider starts to work against the forward movement is represented by the maximum of the pitch rate. From this point the pitch is slowed down until the rider pushes him/herself back to the initial position ( negative pitch rate).

Figure 6. Relative movement of the rider during an autonomous braking intervention (Merkel, 2022).

In preparatory experiments (Merkel & Winner, 2019) the head movement appeared to be the most promising measurement to evaluate the rider adaption to the decelerating motorcycle. This assumption was based on the fact that the head movement was the best-suited measurement for the differentiation between autonomous and manual braking maneuvers (see Figure 7). However, this analysis was limited by the fact that it was based on data from only one rider.

Figure 7. Pitch movement during braking maneuvers: shoulder vs. head (Merkel & Winner, 2020).

The actual second participant study was conducted with 14 participants (12 valid maneuvers). Assessing the data from these experiments, it turns out that comparing different riders, the reaction to an automatic braking intervention is very homogenous for the upper body (measured at the shoulder and lumbar spine level), whereas the head movement can differ significantly (Merkel & Winner, 2020, see Figure 8).

Figure 8. Body pitch rates for all participants after the beginning of an autonomous deceleration (Merkel & Winner, 2020).

The maximum of the pitch rate (rider starts counteracting the forward movement) was analyzed for shoulder, lumbar spine level and head for all riders. For the analysis, the time at which the maximum is reached is particularly interesting. The absolute values are subject to a lot of influencing factors (e.g., body measurements of the rider), so they are barely comparable. Time t = 0 represents the beginning of the brake pressure increase.

As Figure 8 shows, the pitch rates at the shoulder and lumbar spine level follow a characteristic behavior, while the head movement can differ. This can be explained by the fact that the cervical spine is the most flexible part of the spine and the rider might for example raise or lower the head during the maneuver to get a better overview of the situation. This partially occurring difference between the back and the head movement can also be found by comparing the time of the maximum pitch rate over all 12 riders. While the mean time of the maximum pitch rate is still very close for shoulder, lumbar spine and head, the standard deviation is significantly higher for the head. It is about twice as high as for shoulder and lumbar spine (see Table 3).

Table 3. Mean time for maximum pitch rate over all riders in the AEB maneuvers (Merkel & Winner, 2020).

	Mean time of max. pitch rate & std. dev. in s	Min. time of max. pitch rate in s	Max. time of max. pitch rate in s
Shoulder	0,32 ± 0,06	0,26	0,43
Lumbar spine	0,31 ± 0,05	0,25	0,41
Head	0,33 ± 0,12	0,02	0,46

The homogeneity of the upper body movements in autonomous braking maneuvers cannot be retrieved in the maneuvers in which the riders had to apply the brakes themselves. In these cases, the pitch rate curves differ a lot between the riders. This can be explained by the fact that in the manual maneuvers, the deceleration profiles vary significantly. For example, some riders build up brake pressure very fast as soon as they notice the deceleration of the target vehicle and then slowly release the brakes again, while others apply the brakes smoothly and observe the situation and increase the deceleration when they get quite close to the target vehicle. Furthermore, the body movement is not a pure reaction to the changing vehicle state anymore. The rider initiates the deceleration consciously and thus can prepare for it, e.g., by building up body tension prior to applying the brakes. Some riders even show a negative pitch rate while applying the brakes, i.e., they straighten up during the maneuver. The diagrams in Figure 9 show the body movements for all participants in the manual braking maneuvers.

Figure 9. Body pitch rates for all participants in the manual braking maneuvers (Merkel & Winner, 2020).

The observations regarding the manual braking maneuvers allow the conclusion that the automatic braking interventions help to make the rider reaction ‘controllable’ or at least predictable. By provoking an unintentional rider reaction, influences of rider individual (conscious) behavior are interrupted and a characteristic process is initiated. The experiments show that while the rider behavior in rider-induced braking maneuvers is quite inhomogeneous, the unintentional reactions to automatic braking interventions follow a certain pattern. This creates confidence that it is possible to estimate rider reactions when designing automatic braking interventions and to use this knowledge to develop autonomous emergency braking systems that can be used at low risk.

While the head movement appeared promising in preparatory investigations as it showed the most significant differentiation between different maneuvers for one individual rider (Merkel & Winner, 2019), the experiments described in this paper show that this measure shows a clear weakness in terms of reliability and reproducibility when comparing various riders. Thus, the head movement cannot be seen as a liable measure to evaluate if a rider has achieved the ready-for-braking state.

It has been shown that the pitch movement of the rider’s upper body is a more reproducible measure to describe the rider reaction during automatic braking interventions. This measure stays within a slim corridor for all evaluated maneuvers and all riders. Due to the low flexibility of the spine between the lumbar spine level and the shoulder level, the pitch rates at both measuring points stay very close. For future studies this creates confidence that one single pitch rate sensor at the back might be sufficient.

An additional conclusion of the study is that the homogeneity of the rider reactions in autonomous braking scenarios is quite promising regarding the possibility to evaluate the controllability of MAEB systems in studies with relatively small numbers of participants and to transfer the results to a larger number of riders.

Summary of the study results

The multi-stage study concept developed as described at the beginning of this paper was completely run through during the here-described research within the framework of various projects. For a prototype system, controllable parameters for different deceleration profiles were determined in an expert study. Subsequently, the parameterized profiles were examined in a participant study with non-expert riders. In this study, the effectiveness and acceptance of the autonomous interventions were assessed. The prioritized (block) profile was applied in another participant study in order to generate a better understanding of the reaction of unprepared riders to unanticipated braking interventions in terms of characteristic patterns and reproducibility.

The controllability limits determined in the expert study were used to set the parameters of the deceleration profiles for the first participant study. Unprepared riders were confronted with the parameterized interventions and the effectiveness (measured by the possible velocity reduction) and acceptance of the interventions were examined. From the acceptance (interventions are never perceived as ‘dangerous’ in the subjective evaluation) and the absence of loss of control, it can be concluded that an early expert study is suitable for determining appropriate parameters for an automatic braking intervention. The experts' assessment of the controllability of the partial braking profiles is confirmed.

From the results of this first participant study, the deceleration profile ‘block’ emerges as the most promising design of autonomous deceleration during the transition phase. The clear distance to the ramp profile in second place with regard to velocity reduction (32.5 km/h for block vs. 21.5 km/h for ramp within the reference period) allows all other profiles to be discarded and the block profile to be exclusively applied for further investigations.

The investigation of the riders’ individual reaction to the intervention in the second participant study shows that the involuntary physical reaction to an automatic braking intervention follows a characteristic pattern across all participants. The reactions stay in such a narrow corridor that the conclusion suggests that the automatic intervention even contributes to making the rider reaction ‘controllable’ or at least predictable. The homogeneous rider reactions observed in this part of the study create confidence that in the future, even small samples of test persons will be sufficient to obtain meaningful results for the investigation or release of MAEB.

The determination of the pitch rate maximum on the driver's upper body as the peak of the almost symmetrical transition phase confirms the result of the first test person study for the block profile. The pitch rate maximum occurs after about half (0.32 ± 0.06 s) of the transition duration determined there (0.57 ± 0.1 s). However, due to the much smaller scatter in the timing, the pitch rate turns out to be a more reliable measure.

Review of the Method and Conclusion³

Motivated by the challenges described at the beginning of the paper regarding the investigation of active interventions by assistance systems on motorcycles, a method was developed that allows to determine controllability limits of a MAEB without exposing test subjects to an increased risk compared to everyday motorcycling. At the same time, the method allows to investigate natural reactions of unprepared riders do not anticipate the emergency braking situation to occur.

The method was successfully applied to a prototypically implemented system (no environment detection, remote-controlled triggering of braking interventions) and proved to be practicable. It could be shown that automatic braking interventions on motorcycles are basically applicable and, within suitable limits, do not add unreasonable hazard (beyond the risk usually assumed for motorcycling) to the rider.

The development and validation of the method has the potential to contribute to the future release of autonomous emergency braking systems for motorcycles. It opens the opportunity to manufacturers to systematically demonstrate that their developed systems are controllable for users and can therefore be used without creating an additional hazard. From the third part of the study (second participant study) follows the assumption that the involuntary rider reaction is channeled into a controllable path by the unexpected intervention. With respect to this finding, it may be sufficient that a simplified participant study follows an expert study. If a certain braking profile has already been specified, the determination of the safety potential can be omitted in the second study. It then only serves to validate the controllability limits determined by the experts and allows to evaluate the acceptance. The controllability evaluation can be carried out, for example, according to the Response Code of Practice. According to this procedure, an 85 % controllability level is proven if in a study with unprepared participants, 20 out of 20 valid runs fulfil the ‘pass’ criteria (Brockmann, 2009).

False positive interventions can also be covered with a similar procedure. If it cannot be ruled out that a MAEB generates false positive interventions, it must be ensured that these are controllable. The determination of the basic controllability limits could also be realized with an expert study, while in the second step, the controllability is assessed according to a procedure described by Neukum (2015).

The measures determined in the third part of the study to describe rider reaction first of all serve to prove that rider behavior during an automatic emergency braking intervention is reproducible. Once the controllability has been proven for a system, the rider's movement does not need to be permanently recorded. The parameterized deceleration profile can be applied to initiate the rider's transition to a ready-to-brake state, and after a conservatively estimated transition time, the deceleration can be increased to a maximum possible level. Nevertheless, it is conceivable that, in the future, interventions could even be adapted to the individual rider by means of real-time monitoring of the rider's state. For example, riders who show particularly short transition times could be given maximum deceleration earlier and an even greater velocity reduction could be achieved. However, it remains to be discussed whether, with already very short transition times (well below 1 s), the additional effort for continuous recording and processing of the rider's state is justified.

Limitations and open research questions³

In the described research, all studies were performed with the same motorcycle. The extent to which the vehicle geometry influences the limits of controllability and the subjectively perceived criticality remains to be clarified in the future (e.g., touring motorbike with upright seating posture and high, wide handlebars vs. sports motorbike with crouched posture and low stub handlebars).

In the expert study, first impressions were gathered on the influence of the initial speed on controllability limits. However, further investigations are needed here with broader coverage of different speed ranges. Especially with regard to urban riding in the (unstable) low velocity range, further investigations are indispensable if a MAEB is also to be used in this area. The same applies to automatic deceleration during cornering. It can be assumed that in certain roll angle ranges automatic decelerations can no longer be controlled, which has already been indicated withing the expert study (Merkel et al., 2018). However, further investigations are necessary here.

In the described tests, the automatic decelerations were maintained in the safe test environment until the vehicle came to a standstill. Under real conditions (e.g., with following traffic), it can be assumed that automatic braking will be terminated at a certain point, e.g., when the risk of collision has been neutralized, or in order to let the rider take full control of the vehicle's longitudinal dynamics. Here, too, it is necessary to investigate how the deceleration is to be degraded. As with the buildup, involuntary rider reactions are to be expected when deceleration is released, which - just as the initiation of braking - must not destabilize the rider-vehicle system. The degradation is also particularly relevant when any cancellation criterion is reached, e.g., if the rider decides to perform an evasion maneuver during the deceleration and reaches a roll angle that is defined as an operational limit. Here, it must in any case be avoided that in an already critical situation (roll angle build-up during an automatic braking maneuver) an additional potential destabilization occurs due to an unsuitable degradation of the deceleration.

The ideal conditions given in the described studies (rider has both hands on the handlebars, attention is on the riding task) are not always given in real traffic. For example, operating navigation or communication systems can lead to one-handed or even freehand riding, the rider could take his eyes off the road or stand on the footrests instead of sitting in the saddle. It is precisely in such situations of distraction that the intervention of an assistance system may become necessary. If an automatic braking intervention is not ruled out by the system in such cases, the controllability must also be ensured in the event of such so-called misuse cases, and thus must be investigated.

In addition to cases of misuse, the rider reaction in cases of false positive interventions also must be considered. As described above, the first reaction to an automatic braking intervention is an involuntary effect following physical laws. Therefore, no difference in the rider reaction is to be expected in the first moment under the same riding conditions (apart from the lack of a potential collision partner). Nevertheless, it must be ensured that the rider does not subsequently take an undesirable action that leads to a critical condition due to a lack of understanding or misinterpretation of the situation. Thus, false positive interventions represent another future research question.

Beyond the investigation of MAEB, future research could deal with the transferability of the presented method to the investigation of other actively intervening assistance systems for motorcycles.

Ethics

Ethical approval for the participant studies was obtained from the Ethics Commission of the Technical University of Darmstadt under the references EK 32/2018 (first participant study) and EK 39/2019 (second participant study).

Acknowledgements

The described research was conducted during my time as a research associate and doctoral candidate at the Institute of Automotive Engineering (FZD) at the Technical University of Darmstadt. I would like to express my gratitude to my supervisor Prof. Dr. rer. nat. Hermann Winner for giving me the opportunity to work on this topic and for providing me with advice and support at all times.

References

Brockmann, M. (2009). Code of Practice for the Design and Evaluation of ADAS. http://www.acea.be/publications/article/code-of-practice-for-the-design-and-evaluation-of-adas, 2009. Access: 17-11-2021.

Merkel, N. L. (2022). Untersuchungen zur Anwendbarkeit Automatischer Notbremssysteme für Motorräder. Dissertation, Technische Universität Darmstadt, Darmstadt, Germany. DOI: https://doi.org/10.26083/tuprints-00021096

Merkel, N. L., Pleß, R., Scheid, K., Winner, H. (2018, October 1-2). Einsatzgrenzen automatischer Notbrems-systeme für motorisierte Zweiräder – eine Expertenstudie. In Institute for Motorcycle Safety e.V. (Ed.), Proceedings of the 12th International Motorcycle Conference 2018, Cologne/Online, Germany, 2018.

Merkel, N. L., Pleß, R., Winner, H., Hammer, T., Schneider, N., Will, S. (2019, June 10-13). Tolerability of Unexpected Autonomous Emergency Braking Maneuvers on Motorcycles – A Method for Experimental Investigation, In: Conference on the Enhances Safety of Vehicles (ESV), Eindhoven, Netherlands, 2019.

Merkel, N. L., Pleß, R.; Winner, H., Hammer, T., Schneider, N., Will, S. (2022). Automatische Notbremssysteme für Motorräder - Abschlussbericht zum Projekt FE 82.0661/2015. Berichte der Bundesanstalt für Straßenwesen, Reihe F: Fahrzeugtechnik (147), Fachverlag NW in der Carl Ed. Schünemann KG, Bremen, 2022.

Merkel, N. L., Winner, H. (2019, September 9-11). Measures for the Evaluation of Riders’ Adaption to the Changing Vehicle State during Autonomous Emergency Braking Maneuvers on Motorcycles, In Symposium on Bicycle and Motorcycle Dynamics, Padua, Italy, 2019.

Merkel, N. L., Winner, H. (2020, September 1 - October 6). Characteristic Rider Reactions to Autonomous Emergency Braking Maneuvers on Motorcycles. In Institute for Motorcycle Safety e.V. (Ed.), Proceedings of the 13th International Motorcycle Conference 2020, Cologne/Online, Germany, 2020.

Neukum, A. (2008). Beherrschbarkeit fehlerhafter Eingriffe in die Fahrzeugquerdynamik. In Kompaß, K. (Ed.): Fahrerassistenz und Aktive Sicherheit, Haus der Technik Fachbuch. expert verlag, Renningen, 2015.

Neukum, A., Lübbeke, T., Krüger, H.-P., Mayser, C.. Steinle, J. (2008). ACC-Stop&Go: Fahrerverhalten an funktionalen Systemgrenzen. In Maurer, M.; Stiller, C. (Eds.): 5. Workshop Fahrerassistenzsysteme - FAS2008. Karlsruhe, Germany, 2008.

Savino, G, Giovannini, F, Baldanzini, N, Pierini, M. & Rizzi, M. (2013). Assessing the potential benefits of the motorcycle autonomous emergency braking using detailed crash reconstructions. In Traffic injury prevention, 14, 40–49. DOI: https://doi.org/10.1080/15389588.2013.803280

Savino, G., Pierini, M., Thompson, J., Fitzharris, M., Lenne, M. G. (2016). Exploratory field trial of motorcycle autonomous emergency braking (MAEB). In Traffic injury prevention (8), 17, 855–862, 2016.

Savino, G., Pierini, M., Baldanzini, N. (2012). Decision logic of an active braking system for powered two wheeler. In Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering (8), 226, 1026–1036, 2012.

Symeonidis, I., Kavadarli, G., Schuller, E., Graw, M., Peldschus, S. (2012). Analysis of the stability of PTW riders in autonomous braking scenarios, In Accident Analysis and prevention,49, 212–222, 2012.

---

1. This section represents the situation at the beginning of the here-described research, several years ago. Up until today, various other research groups have of course also proceeded in their MAEB research.↩︎

2. Large parts of the text in this section have already been published in Merkel & Winner (2020). Nevertheless, it is important to repeat it here, as the results are an integral part in the overall context of the here-described investigation method. Permission to use was granted by the editors.↩︎

3. Contents of the review and limitations sections are translated from Merkel (2022). The source is licensed under the CC BY-SA 4.0 license.↩︎