Assessing the Potential Benefits of the Motorcycle Autonomous Emergency Braking Using Detailed Crash Reconstructions

Abstract

Objective: The aim of this study was to assess the feasibility and quantitative potential benefits of a motorcycle autonomous emergency braking (MAEB) system in fatal rear-end crashes. A further aim was to identify possible criticalities of this safety system in the field of powered 2-wheelers (PTWs; e.g., any additional risk introduced by the system itself).

Methods: Seven relevant cases from the Swedish national in-depth fatal crash database were selected. All crashes involved car-following in which a non-anti-lock braking system (ABS)-equipped motorcycle was the bullet vehicle. Those crashes were reconstructed in a virtual environment with Prescan, simulating the road scenario, the vehicles involved, their precrash trajectories, ABS, and, alternatively, MAEB. The MAEB chosen as reference for the investigation was developed within the European Commission–funded Powered Two-Wheeler Integrated Safety (PISa) project and further detailed in later studies, with the addition of the ABS functionality. The boundary conditions of each simulation varied within a range compatible with the uncertainty of the in-depth data and also included a range of possible rider behaviors including the actual one. The benefits of the MAEB were evaluated by comparing the simulated impact speed in each configuration (no ABS/MAEB, ABS only, MAEB).

Results: The MAEB proved to be beneficial in a large number of cases. When applicable, the benefits of the system were in line with the expected values. When not applicable, there was no clear evidence of an increased risk for the rider due to the system.

Discussion and Limitations: MAEB represents an innovative safety device in the field of PTWs, and the feasibility of such a system was investigated with promising results. Nevertheless, this technology is not mature yet for PTW application. Research in the field of passenger cars does not directly apply to PTWs because the activation logic of a braking system is more challenging on PTWs. The deployment of an autonomous deceleration would affect the vehicle dynamics, thus requesting an additional control action of the rider to keep the vehicle stable. In addition, the potential effectiveness of the MAEB should be investigated on a wider set of crash scenarios in order also to avoid false triggering of the autonomous braking.

Supplemental materials are available for this article. Go to the publisher's online edition of Traffic Injury Prevention to view the supplemental file.

Keywords:

• autonomous braking
• collision mitigation
• crash
• effectiveness
• injury
• motorcyclist

Introduction

Today motorcycles can represent important alternative means of transportation, due to the growing congestion in urban areas and the demand for more time- and energy-efficient transports. However, previous research has shown that in the case of a crash, motorcycle riders have much higher injury risks than passenger car occupants (Gabler 2007). Although innovative protective equipment for motorcyclists (i.e., air bag jackets) has been developed with promising results (e.g., Pellari 2012), impact speeds are often too high to prevent severe injuries (Rizzi et al. 2009).

The approach for reducing injureis due to road crashes is mostly known in the automotive safety area as integrated safety; the precrash and crash phases (Haddon 1980) are no longer considered separately but rather interact with each other (Strandroth 2012). With this approach the entire chain of events leading to a crash can be understood and possible countermeasures can be analyzed in order to either prevent crashes or mitigate injuries when the crash is no longer avoidable (Strandroth 2012).

Fig. 1 Prototype vehicle of the PISa project equipped with MAEB (inset: laser scanner mounted on the front fairing).

Following this approach, the development and implementation of autonomous emergency braking (AEB) on passenger cars is ongoing (European New Car Assessment Programme 2012). Most of these systems are designed to avoid and mitigate car-to-car rear-end crashes and/or crashes with pedestrians and cyclists. Assessments based on real-life crashes in the United States and Sweden have shown a 25 percent reduction of rear-end crashes (Highway Loss Data Institute 2011; Isaksson-Hellman and Lindman 2012). In addition, autonomous emergency braking can enhance the benefits of other crash protection systems (i.e., air bags for vulnerable road users [VRU]) by reducing collision speeds (Fredriksson and Rosén 2012), which is the key point of integrated safety.

Though the same approach is feasible for motorcycle safety, few safety systems have been introduced in this field. Today the most common one is the antilock braking system (ABS), which was first introduced in the late 1980s. ABS has been shown to generally provide shorter stopping distances (Green 2006) and can also prevent the motorcyclist from falling to the ground due to increased braking stability (Teoh 2011). Evaluations in real-life conditions have shown great benefits, with crash reductions ranging from 22 to 37 percent (Highway Loss Data Institute 2009; Teoh 2011).

A further step toward an integrated safety for motorcycles was made in 2009 when a first prototype of motorcycle autonomous emergency braking (MAEB) was developed in the Powered Two-Wheeler Integrated Safety (PISa) project (Grant et al. 2008).

This system can detect an obstacle through a laser scanner located in the front of the motorcycle (Figure 1), warn the rider, and brake automatically to produce a 0.3 g medium deceleration (autonomous braking, AB) if the rider does not react before the collision becomes unavoidable. However, if the rider does brake after this point, maximum braking force is automatically applied (enhanced braking, EB; Savino, Pierini, and Baldanzini 2012).

In the present research, the PISa autonomous braking system was modeled and incorporated a standard ABS, which also operated when the AB was not triggered. The characteristics of the MAEB are summarized in Table 1.

Table 1 Characteristics of the MAEB

Description	Value
Detection range	200 m
Detection angle	100°
AB reference deceleration	3 m/s^2, 9 m/s^2 (EB)
Minimum obstacle width	1 m

Table 2 Description of the analyzed cases

Case	Make, model, model year	Speed limit (km/h)	Motorcycle braking before crash	Brake skid (m)	Initial speed (km/h)	Collision speed (km/h)	Impact configuration
A	Triumph Thunderbird ’03	90	No	—	90	90	Upright
B	Suzuki DL1000 ’06	70	Yes	13	100	65–70	Upright
C	Suzuki GSX750 ’83	70	Yes	17	80	45–50	Fell on right-hand side before crash
D	Yamaha Teneré 750 ’98	50	Yes	0	75	60	Upright, with roll angle
E	Muz Scorpion S ’97	70	Yes	0	70	50	Upright, with small roll angle
F	Yamaha R6 ’99	70	Yes	0	105	70–75	Upright
G	Suzuki Intruder 1500 ’98	70	Yes	20	70	45–50	Upright

Previous studies have retrospectively evaluated the potential of MAEB in real-life crashes through simple analysis of in-depth material (Rizzi 2011; Savino, Pierini, Rizzi, and Framptom 2012; Strandroth 2012), showing promising results. Further research is needed to understand more deeply the true benefits of MAEB in different crash scenarios. Therefore, the present study focused on a specific crash type as an additional step to evaluate MAEB. A sensitivity analysis of the potential benefits of the MAEB was conducted taking into account factors such as road scenario, dynamical state of the vehicles involved, and the rider's control actions. Therefore, the use of advanced simulation software was considered more suitable than a method based on engineering judgments by experts.

Study Objectives

In summary, the objectives of the study are the following:

•	further assess the feasibility and the quantitative potential benefits of the MAEB with special focus on fatal rear-end crashes;
•	identify possible criticalities of the MAEB in the field of powered 2-wheelers (PTWs) in fatal rear-end crashes; for example, any additional risk to the rider due to the system itself.

Methodology

Material

The present study used in-depth studies of fatal crashes collected by the Swedish Transport Administration (STA). The STA has been carrying out in-depth studies for each fatal road crash since 1997. Crash investigators at the STA systematically inspect the vehicles involved in fatal crashes and record direction of impact, vehicular deformation, air bag deployment, tire properties, etc. The crash site is also inspected to investigate road characteristics, collision objects, braking skids, etc. Further information about injuries and use of helmets and protective equipment is provided by forensic examinations, questioning and witness statements from the police, and reports from the emergency services. The chain of events leading to the fatal crash is analyzed and critical events are identified; collision speeds are generally derived by vehicular deformation, and initial driving speeds are mostly based on witness accounts, brake skids, etc.

Because all fatal crashes are included in the sampling criterion, the material is fully representative for Swedish conditions and possibly for northern countries generally.

Case Selection and Description

This study focused on rear-end crashes in which the motorcycle was the bullet vehicle. A typical car-following scenario can be described as the motorcycle traveling on a straight path when the lead vehicle (traveling in the same direction) slows down or stops; that is, to turn into a lateral road or because of other traffic ahead. Only crashes in which the motorcyclist sustained fatal injuries in the rear-end collision itself were included; in other words, crashes where the fatal injuries occurred in further secondary collisions (i.e., other than the initial impact with the lead vehicle) were excluded from the analysis. In total, 7 fatal crashes involving car-following were identified for the period 2006–2009 in Sweden. During the same period, 213 motorcyclists were killed according to Swedish official statistics (Swedish Transport Agency 2013).

The time of the precipitating event was defined in general terms as the time at which the major event that led to the crash took place, in addition to other secondary, although relevant, contributory events. The above definition allowed identifying a specific precipitating event for each case (e.g., case E: The lead vehicle [passenger car] stopped at an intersection to turn into another rural road; case F: the lead vehicle [passenger car] made a U-turn). Reaction times, braking distances, and other factors (i.e., swerving, driving directions and speeds, etc.) were then derived starting from the precipitating event for each of the 7 relevant cases mentioned above. All crashes occurred in daylight, with clear sky and dry asphalt. All riders used helmets and were not under the influence of alcohol or illegal drugs. No ABS-equipped motorcycles were involved in these crashes. A brief description of these cases is given below and a summary of the cases is provided in Table 2.

Case A

The lead vehicle (passenger car) slowed down and stopped at an intersection to turn left into another rural road (precipitating event). While waiting for 7 oncoming motorcycles, the passenger car was hit from behind by a motorcycle. The rider (male, 66) sustained fatal internal bleeding. According to witness accounts, the killed motorcycle rider may have been distracted by the oncoming group of motorcycles.

Case B

The lead vehicle (light truck) stopped to turn left into a secondary rural road (precipitating event) and waited for 6 oncoming motorcycles. A motorcycle traveling along the lane of the truck hit the rear of the truck. The rider (male, 44) sustained fatal thorax injuries; he might have been distracted by the oncoming group of motorcycles.

Case C

The lead vehicle (light truck) was about to turn left into a private road (precipitating event). A motorcyclist (male, 71) coming from behind may have understood this too late and attempted to swerve while braking. He lost control and slid into the side of the truck, sustaining fatal thorax injuries.

Case D

The lead vehicle (passenger car with trailer) was about to turn left into the driveway of a private property (precipitating event). A motorcyclist (male, 39) collided from behind into the trailer's left side. He sustained fatal thorax injuries.

Case E

The lead vehicle (passenger car) stopped at an intersection to turn into another rural road (precipitating event) and waited for some oncoming motorcycles. A motorcycle with 2 riders (males, 71 and 14) following from behind attempted to swerve and crashed into the right rear of the car. The motorcycle driver sustained fatal head and thorax injuries.

Case F

After turning the wrong way out of a roundabout in an urban area, the lead vehicle (passenger car) made a U-turn (precipitating event). A motorcycle following from behind hit the left side of the car. The rider (male, 21) sustained fatal head injuries.

Case G

The lead vehicle (passenger car) was following another car and a bus on a rural road. When the bus slowed down before a bus stop (precipitating event), the 2 following cars braked, too. The motorcyclist (male, 77) overbraked and partly slid into the rear of the lead vehicle, sustaining multiple fatal injuries.

Simulations

Part 1. MAEB vs. Nothing

The 7 crashes were reconstructed with the numerical tool PreScan (Gietelink et al. 2004) in order to perform a sensitivity analysis of each case. Such an analysis was intended to provide (a) a more robust data set for the quantitative assessment of the MAEB performance compared to a single-shot reconstruction and (b) a preliminary investigation of the importance of various factors that have an influence on the crash outcome.

Each case was characterized in detail by the following information:

•	road scenario: in-scale geometry of the road at the crash site, reconstructed from satellite maps, and type of infrastructure, whenever of relevance;
•	vehicles involved: type and their initial state at precipitating event in terms of position, heading, speed, acceleration, and maneuvers executed during the event.

The motorcycles used in the simulations were general models reasonably similar to the actual vehicles. The realistic dynamic behavior of the motorcycles in the simulations was obtained by using a detailed multibody vehicle model generated by the simulation tool BikeSim (Mechanical Simulation, Ann Arbor, MI) and coupled with PreScan (TASS International, Helmond, The Netherlands) in a MATLAB Simulink environment.

A design of experiment (DOE; Fisher 1935) of 3 parameters varying on 3 levels, using a full-factorial scheme, was applied to each case in order to thoroughly explore variations in the nominal cases and to quantitatively assess the influence of the parameters on the speed reduction. A 3-level DOE was chosen due the expected nonlinear relation of the parameters with the speed reduction produced by MAEB. The generated sets of crash scenarios were simulated both in real conditions (with no assistance system installed on the motorcycle) and with the MAEB. The functionalities of the MAEB were described in Savino, Pierini, and Baldanzini (2012).

The parameters used in the DOE were the initial PTW lateral position on the road, the initial PTW speed, and the rider's reaction time. Quantification of the latter parameter was done considering that in nominal conditions (i.e., in the situation of the crash reproduced using the original in-depth information) the rider maneuver (in steady-state conditions) was initiated 0.8 s after the perception of the impending crash. The value of 0.8 s was selected according to Roll et al. (2009), who indicated 0.7–0.9 s as the average time interval between the time step of perception of the impending crash and the time step when the evasive maneuver is in steady-state conditions (e.g., full braking). In the case of early or late reaction, the rider maneuver is anticipated/delayed by 0.5 s, thus representing the extreme amplitude values of a rider's reaction time to perform the maneuver (Davoodi et al. 2011; Roll et al. 2009). The variable levels used in the DOE are reported in Table 3. In all cases the output parameter was the impact speed, because it is related to the energy dissipated in the impact and to the rider's injuries.

Table 3 Parameters and associated levels used in the DOE for each crash (percentages are for the nominal value of the variable)

Parameter	−1/0/1 level
Initial PTW lateral position	−0.5/0/+0.5 (m)
Initial PTW speed	80/100/120 (%)
Rider reaction timing	0.3/0.8/1.3 (s)

A total of 54 simulations (3 variables, 3 values per variable, 2 systems) were performed for each case.

Table 4 Assumed boundary conditions

Case	Trajectory PTW	Opponent vehicle maneuver	Host PTW maneuver
A	Straight	Still	None
B	Straight	Still	Hard braking (80f, 60r [bar])
C	Straight	Moving (left turn into lateral street)	Hard braking (80f, 80r [bar]) plus right swerving (10° target roll)
D	Impact point out of a slight curve	With trailer, moving (left turn into lateral street)	Poor braking (40f, 40r [bar]) plus left swerving (−10° target roll)
E	Straight	Still	Medium braking (30f, 30r [bar]) plus right swerving (10° target roll)
F	Straight	Performing U-turn	Medium braking (50f, 0r [bar])
G	Straight	Decelerating until stopped	Medium braking (40f, 60r [bar])
f and r, reported in the last column, refer respectively to the front and rear brakes.

Part 2. MAEB (With and Without EB) vs. ABS

In addition to the above-mentioned assessment, 2 other comparisons were performed: MAEB vs. ABS and pure emergency braking function (i.e., MAEB with enhanced braking function deactivated) vs. ABS. In these comparisons, the full set of crashes was considered in nominal conditions and the variability was introduced only on the rider behavior. In addition to the nominal maneuver, the rider was considered able to perform the following:

•	panic heavy braking: 80 bar pressure applied to master cylinder on both the front and rear brakes;
•	poor braking: 40 bar pressure applied to master cylinder on both the front and rear brakes.

The above-mentioned pressure values were derived from experimental trials conducted with fully instrumented PTWs, which allowed a linking of the deceleration with the brake pressures. In particular, the authors chose 80 and 40 bar as reference pressures for panic braking and poor braking, respectively, in accordance with the experimental data extracted from the database of acquisitions in naturalistic riding conditions collected within the European Commission–funded 2BeSafe project by a team at the University of Florence (Weare et al. 2011). In case a composite avoidance maneuver resulted from the in-depth database (i.e., swerve plus braking), the swerve action was still coupled to the braking behaviors previously described. The latter 2 braking avoidance maneuvers are typical of PTW–car crashes (Penumaka et al., in press).

The assumed boundary nominal conditions are summarized in Table 4. Nine simulations (3 braking behaviors, 3 systems) were performed for each case.

Table 5 Summary of the results for the 7 cases in the actual configurations

Case	Impact speed (m/s)	Impact speed reduction (m/s)	MAEB activation mode
A	24.9	1.8	Pure AB
B	20.3	0.2	EB
C	13.2	Crash avoided	ABS
D	15.2	0.1	ABS
E	14.4	1.0	EB
F	18.0	1.3	EB
G	13.7	2.9	EB

Results

A total of 378 simulations were run for part 1 and 84 were run for part 2. The results of part 1 (MAEB vs. nothing) and part 2 (MAEB vs. ABS) are presented in the following paragraphs.

Part 1. MAEB vs. Nothing

Actual Configurations

Starting from the actual configurations of the crash cases, the results of the simulations showed that the MAEB deployed in 5 out of 7 cases: cases A, B, E, F, and G. In case A where the rider did not attempt any braking, the pure AB functionality was activated, producing an impact speed reduction of 1.8 m/s. In the other 4 cases, the enhanced braking was triggered due to the rider's braking action and the final impact speed reductions were, respectively, 0.2, 1.0, 1.9, and 2.9 m/s.

In cases C and D, the MAEB did not deploy due to the inhibition given by the roll angle produced by the swerving attempt. In those cases, the ABS functionality was engaged, producing complete collision avoidance in case C and an impact speed reduction of 0.1 m/s in case D. A synthesis of the results in the actual configurations is provided in Table 5. A diagram showing the impact speed vs. the impact speed reduction is given in Figure 2. In the diagram, the theoretical impact speed reduction is plotted in case of a fixed obstacle and (a) no rider reaction and (b) medium braking applied at the instant of MAEB triggering.

Modified Configurations

In most of the crash cases included in the study, the collision with the MAEB off did not take place in the modified configurations at lower speed and early reaction timing (LSER configurations; cases B, D, E, and F).

Case A

In all 26 modified configurations the collision took place. The MAEB was triggered in pure AB mode, producing an impact speed reduction ranging from 1.4 m/s (at higher speed and with lateral offset on either side) to 1.8 m/s (actual timing, higher speed, and actual lateral position).

Fig. 2 Impact speed reduction (m/s) due to MAEB in the actual configuration. Theoretical speed reduction curves considering no braking and medium braking (6 m/s²).

Case B

Concerning case B, the collision with MAEB turned off did not take place in the LSER configurations only.

When the MAEB was on, in those configurations where the collision took place the EB was triggered, producing an impact speed reduction ranging from 0.2 m/s (actual configuration) to 2.9 m/s (late reaction, higher speed, lateral offset on the left-hand side).

Case C

In 9 out of 26 modified configurations, the PTW collided with the lead vehicle in an upright position (in 2 of those cases with no roll angle). In the remaining configurations, the PTW fell down before the impact, similar to the actual configuration.

The AB or EB did not trigger in any of the modified configurations with the MAEB on. The ABS avoided the crash in 17 out of 26 configurations. In those configurations where the collision took place, the impact speed reduction due to the ABS ranged from 0.0 m/s (actual timing, higher speed, and actual lateral position) to 1.5 m/s (late reaction, higher speed, actual lateral position).

Case D

The collision with the MAEB turned off did not take place in the LSER configurations only.

When the MAEB was on, the triggering of the EB took place in 3 configurations: higher speed and late reaction timing (irrespectively of the lateral position). In those configurations, the impact speed reduction ranged from 1.2 m/s (lateral offset on the right-hand side) to 2.3 m/s (actual lateral positioning).

The ABS functionality avoided the crash in 3 configurations with early reaction of the rider, 2 of them at the actual speed and one at higher speed. The impact speed reduction of the ABS ranged from 0 up to 2.3 m/s (late reaction, actual speed, lateral offset on the left-hand side).

Case E

With the MAEB turned off the collision did not take place in the LSER configurations only.

In the remaining configurations with the MAEB on, the EB triggered in 16 simulations, corresponding to all of the configurations with a late reaction and all of the configurations with actual timing except for those at low speeds. The impact speed reduction ranged from 0.4 m/s (current timing, lower speed, actual position) to 2.3 m/s (later reaction, lower speed).

The ABS functionality could avoid 2 collisions and in the remaining configurations the impact speed reduction ranged from 0 to 0.9 m/s (actual timing, lower speed, lateral offset on the right-hand side).

Case F

With the MAEB turned off the collision did not take place in the LSER configurations only.

In the remaining configurations with the MAEB on, the EB triggered in all the configurations and the impact speed reduction ranged from 0.1 m/s (late reaction, lower speed, offset on the right-hand side) to 6.3 m/s (early reaction timing, higher speed, offset on the right-hand side).

Case G

With the MAEB off the collision did not take place in the LSER configurations only.

In all of the remaining configurations when the MAEB was on, the EB was triggered, producing an impact speed reduction in the range between 1.2 m/s (late reaction, higher speed, and lateral offset on the left-hand side) and 5.1 m/s (early reaction, actual speed, actual lateral position).

DOE Analysis

For each case, a DOE analysis was performed to investigate the main effects of the 3 experimental variables (reaction timing, speed, and lateral position). The main effects of cases A, B, F, and G, in which the AB or EB functionality was largely involved, are reported in Figure 3. The plots of the remaining cases and of the interactions are included in the Appendix (see online supplement).

Fig. 3 Analysis of effects of experiment variables on mean impact speed reduction (m/s) due to MAEB for cases A, B, F, and G.

A diagram showing the impact speed vs. the impact speed reduction for all configurations including the actual ones, neglecting those where the impact was completely avoided, is provided in Figure 4. In the diagram, the theoretical impact speed reduction is plotted in the case of a fixed obstacle and (a) no rider reaction and (b) medium braking applied at the instant of MAEB triggering.

Fig. 4 Impact speed reduction (m/s) due to MAEB (ABS + AB + EB) in the actual configuration. Theoretical speed reduction curves considering no braking and medium braking (6 m/s²) are indicated.

Fig. 5 Impact speed reduction due to MAEB in comparison with ABS alone, considering 3 braking actions: actual, panic, and poor (color figure available online).

Part 2. MAEB (With and Without EB) vs. ABS

In all of the nominal conditions and in all of the additional configurations obtained with different braking actions (actual, panic, and poor braking), the impact speed with the MAEB on was equal or lower than that with ABS alone. Diagrams showing the impact speed reduction due to the MAEB in comparison with the ABS are provided in Figure 5. A distinction is provided for the pure AB functionality (MAEB without EB functionality) and the full MAEB.

The simulations also showed that ABS alone would have avoided the crash in case C for any of the 3 analyzed rider braking actions, and neither ABS nor MAEB would have significantly affected the outcomes for case D.

Discussion

The MAEB, consisting of ABS, AB, and EB functionalities, was relevant in all 7 car-following crashes considered in the study and the AB/EB was triggered in 5 cases. The MAEB is supposed to trigger once the collision becomes physically unavoidable either by braking or swerving. Because at high speeds the swerve maneuver is more efficient than braking (Giovannini et al. 2013), the MAEB activation is delayed until even a swerving action would not avoid the crash, thus limiting the potential impact speed reduction with respect to lower speed values. The simulations indicated an impact speed reduction due to MAEB compatible with the theoretical values in 4 out of 7 cases (Figure 2). In a fifth case, where the AB/EB was triggered (case B), the benefits were minor mainly due to the hard braking action of the rider associated with an early reaction (as the crash reconstruction indicated).

In the 2 cases in which the AB/EB did not trigger (cases C and D), the system deployment was inhibited by the rider's swerving action. In case C the crash was avoided due to the ABS functionality. In another case where the rider tried to swerve (case E), the AB/EB first triggered and then turned off due to the roll angle of the PTW, producing impact speed reduction. In none of the cases did the presence of MAEB and its intervention on the vehicle dynamics produce evidence of additional threat. No fall events were predicted with the MAEB turned on.

Concerning the modified configurations, the effects of the MAEB in terms of impact speed reduction, positive and negative variations with respect to the effects of the MAEB in nominal conditions were reported depending on the case and the modified variables. The range increased from 0–3 m/s in the actual configurations to 0–4 m/s in the modified configurations. Globally the values were in the range of those estimated with the theoretical model (Figure 4). Though the speed reduction range in the modified configuration was similar to the range in the actual configurations, the analysis of the main effects (Figure 3) demonstrates the strong influence of parameter variation on the MAEB speed reduction.

Analysis of the interaction plots showed a considerable mutual influence only between timing and speed for the cases B, C, F, and G, and relevant interactions among all parameters are reported for the remaining cases.

The results of part 2 of the study (MAEB vs. ABS) showed the influence of the AB/EB and pure AB on the impact speed reduction operated by the MAEB in comparison with ABS functionality alone. In case A, where the rider did not react before the impact, the effects were only due to the pure AB. In the remaining cases, the pure AB did not produce a relevant speed reduction compared to the ABS. In case B, in the actual configuration, the full MAEB slightly influenced the impact speed and the same result was obtained in the panic braking modified configuration. In fact, the rider performed a hard braking maneuver. When considering the modified configuration of poor braking, the MAEB would be more effective in terms of impact speed mitigation. Cases C and D were only affected by the ABS as previously mentioned and the MAEB did not deploy in the modified configurations of panic braking and poor braking. Cases E, F, and G were influenced mainly by the EB functionality. For those 3 cases, the panic braking configuration would reduce the effectiveness of the EB in terms of speed reduction, in particular in case E. However, in modified configurations of poor braking, the effects of MAEB would be higher in case E with respect to the actual configuration.

This study confirms the applicability of the MAEB in real crash cases in car-following scenarios, showing possible impact speed reductions in a range of initial conditions. The modified initial conditions influenced the final speed reduction operated by the MAEB, but no risks associated with the deployment of the MAEB were evidenced. Specifically, in all of the analyzed configurations, the MAEB never increased the impact speed with respect to the vehicle without the system installed. Moreover, the system activation did not influence the motorcycle's stability and maneuverability (no falling events were reported in any of the modified configurations when the MAEB was on). The main effects of the different configurations on the effectiveness of MAEB confirmed the nonlinear nature of the phenomenon (Figure 3); however, it is impossible to identify a unique trend in the data for different scenarios, which can lead to generalization of the results. Because each crash configuration is a unique combination of parameters and their mutual influence, future validation studies should use a random sampling scheme to explore the domain of nominal and modified configurations. In addition, to confirm the robustness of the MAEB behavior, a wider data set of accidents, capable of covering the conditions of car-following accidents, is a necessary requisite, because no interpolation or extrapolation of the data is feasible.

The results also showed the positive effects of ABS, although relevant speed reductions were achieved both by AB (in the case of no reaction) and EB (when the rider operated the brakes). Moreover, the positive effects in terms of speed reduction were more evident for poor braking actions than in panic hard braking conditions. The latter results suggest that the following 2 aspects should be investigated in order to increase the safety potential of the MAEB:

•	the effectiveness of the AB functionality to act as warning signal for the rider to stimulate her or his reaction (i.e., braking or swerving);
•	the study and possible improvement of the definition of the AB deceleration value, based on real-world data.

The present study also illustrates a critical aspect of the MAEB in car-following crash scenarios; that is, the possibility of a swerve attempt by the rider prior to the collision. In case of swerving, the MAEB is supposed to be inhibited to avoid vehicle destabilization, according to the decision logic of the system (Savino, Pierini, and Baldanzini 2012). Cases C and D showed unsuccessful swerving attempts where the roll angle values above the threshold of 10° inhibited MAEB triggering. As a consequence, in those cases the potential speed reduction due to the EB functionality would not have been exploited, even if the roll angle was slightly above the threshold. Further consideration should be given in the future to redesigning the inhibition criteria in order for ineffective swerve attempts to potentially benefit from MAEB.

Limitations

Though the results of the present study seem promising, caution is needed due to a number of limitations. First, only fatal crashes from Sweden were included in this study. The analyzed cases were a fully representative sample for Sweden, and possibly for Northern Europe; however, it could be argued that motorcycling is somewhat different in Sweden than in other countries due to climate. Therefore, the analyzed crashes could differ from crashes in other European areas.

General caution is also needed when only fatal crashes are analyzed because crashes with lower injury outcomes would probably differ in terms of riding and impact speed ranges. Furthermore, the sequence of events leading to a fatal motorcycle crash is probably different from those with milder injury outcomes, which means that crash and injury mechanisms (braking vs. no braking, upright vs. prone, etc.) would probably be different as well. Future research should address this issue.

A further limitation is due to the material source. Though previous studies have shown that the STA's material is sufficiently reliable for in-depth analysis (Strandroth 2012), it should be noted that most of the input values were based on postcrash investigations and not on real-time measurements. For instance, collision speeds were not measured with crash pulse recorders; vehicle trajectories and driver reactions were not recorded either. Though input errors would clearly affect the simulations and the overall results, an attempt to control for this uncertainty was made by analyzing the variation of collision speed depending on 3 of the most important parameters (initial PTW lateral position, speed, and rider reaction times). The results showed a reasonable variation in collision speeds, thus suggesting that the simulations were robust and consistent.

Finally, with the current state of knowledge the expected values of impact speed reduction due to MAEB cannot be correlated with a reduction in the injury outcomes for the rider. In the future, the creation of specific motorcycle rider risk curves, similar to those created for passenger cars, will allow estimation of the potential benefits of MAEB in terms of injury reduction in car-following scenarios.

Conclusions

This study presented an investigation of the effectiveness of MAEB applied to virtual reconstructions of 7 fatal PTW crashes that occurred in Sweden and involved a car-following crash scenario. The simulations were performed with and without the MAEB system on the PTW in order to investigate the potential benefits of this safety device in terms of impact speed reduction. A DOE analysis was performed to identify possible performance trends of the MAEB in the car-following scenarios. The simulations used for the DOE analysis were also employed to assess the robustness of the MAEB system. A further analysis was performed to compare the performances of the ABS and MAEB systems on the same set of nominal configurations, taking into account different rider braking behaviors.

The results confirmed the general applicability and effectiveness of MAEB in car-following scenarios: the MAEB was triggered in 5 out of 7 cases and produced a maximum speed reduction of 4 m/s (considering all simulations). In addition, the data showed no negative influence of the MAEB system on the vehicle stability, because the PTW did not fall before the collision in any of the analyzed configurations when the MAEB was on.

The MAEB did not trigger (2 out of 7 cases) when the emergency brake activation was inhibited due to the emergency swerving maneuver performed by the rider prior to collision. The MAEB inhibition, in the case of unsuccessful swerving maneuvers by the rider, strongly limits the effectiveness of the system. An improvement of the inhibition criteria and their refinement is a critical aspect of the safety system. Future studies should investigate the interaction among the AB functionality, swerving maneuvers, and riders’ behaviors in order to limit the inhibition of the system only in the case of an avoidable collision (by any evasive maneuver).

The DOE analysis showed the uniqueness of each accident crash scenario and no general trends could be identified within the data set of fatal car-following crashes. Future studies should apply a random sampling scheme over the variable range of the initial conditions to test the robustness of the MAEB system.

The results of the simulations conducted with simple ABS, pure AB, and full MAEB considering 3 rider braking behaviors showed that the EB functionality could reduce the impact speed with respect to ABS (the effects were more evident in case of poor braking maneuvers), whereas the pure AB functionality affected only the case where the rider did not brake prior to collision.

In conclusion, the present work has proposed a methodology for virtual testing of PTW integrated safety devices using real crash scenarios. The methodology has demonstrated that MAEB is a potential candidate device to enhance rider safety in car-following scenarios. In order to confirm the validity of these preliminary results, the assessment should be repeated on a more extended data set, because no interpolation or extrapolation of the analyzed scenarios can be performed.

Acknowledgments

Many thanks to Johan Strandroth at the Swedish Transport Administration for providing access to the in-depth data. This research activity was partially funded by the Province of Prato within the project Road Safety of Powered Two-Wheelers.

© Giovanni Savino, Federico Giovannini, Niccolò Baldanzini, Marco Pierini, and Matteo Rizzi