Analysis of ACC-Concepts with the simulation tool PELOPS

Dr-Ing., Dipl.-Wirt.-Ing. Jens Ludmann
Institut für Kraftfahrwesen

---

Introduction
ACC-Concepts and simulation requirements
The ACC-simulation in PELOPS
Analysis of ACC-systems
Conclusion

---

Introduction

After more than 100 years of vehicle development, we now have a very high technical standard in the automotive industry. This long periode of development was mainly related to the vehicle itself, which means an isolated view of this transport system. Nowadays there is a new period in the vehicle development, which extends the former subjects about the connection between the vehicle and it's environment, as there are infrastructure and traffic related aspects (fig. 1)



Fig. 1: Driver assistance concepts

The aim of this extension is to assist the driver by improoving the safety, to find the destination in unknown areas, to aviod time losses in jam's and to improove the driving comfort as well as an improovement of the traffic efficency. One of these systems is the Adaptive Cruise Controll (ACC), which adapts the longitudinal vehicle controll to the vehicle ahead.

ACC-Concepts and simulation requirements

The basic research on ACC was executed in coordination between several automotive companies in PROMETHEUS. At the end of PROMETHEUS ACC reached a high level of development. At present the developed for production is running. In Europe the production start can be considered for the next one or two years. In Japan there is already an ACC-vehicle on the market (fig. 2)



Fig. 2: First ACC-vehicle in production: Mitsubishi DIAMANTE

To regulate the speed and distance to a leading vehicle the system controls the engine power by throttle and shifts down the automatic transmission for decelleration, if the negative torque has to be enlarged. The sensor system of this ACC-vehicle is based on a laser radar installed into the fron bumper and a video camera installed in front of the rear mirror. The laser radar detects distance and speed difference to previous cars, and the video camera recognizes the running lane by image processing. With the combination of both the distance and speed difference to the relevant vehicle is created. Because of the low capabilities for decelleration the desired distance of the system is set to 2 sec.. The average distance on german motorways with high traffic load is in the range of 1,0s to 1,2s.. These low distances can not be controlled without aktive braking, so that this concept is not useful for european countries. The following Figure (fig.3) shows differences in ACC- concepts.



Fig. 3: ACC-Concepts

The difference between the concepts ist the arrangement of actuators, which convert the ACC-strategy into longitudinal control. There are the throttel or IC-controll (ac1), the brake actuator (ac2) and the access on the gearbox (ac3). With these three actuators different ACC- concepts can be designed. One concept is a combinition of all actuators (ac1, ac2 and ac3) which gives the highest functionality of an ACC-system. But this concept is very expensive. The cheapest solution uses only the throttel or IC-controll (ac1), but the functionality of such a system is very poor because the decelleration is too low for the normal traffic. It looks as if concepts for the european countries will include the throttel or IC-controll (ac1) and the brake actuator (ac2). For the USA and Japan a concept without active breaking (ac1 and ac3), like the Mitsubishi Diamante, seems to be a reasonable solution. Independent of the actuator combination it is neccessary to model all possible ACC-components in deatil for any analysis of ACC-systems. For a simulation program this requires a detailed vehicle model and all traffic components as there are the stretch, the environment and the traffic load. The vehicle model must include the driveline (including it's nonlinearity), an automatic gearbox and a module for the throttle and the brake (fig.4).

simulation models

vehicle model	module for the throttle module for the brake automatic gearbox driveline
traffic components	stretch environment traffic load

Fig. 4: Relevant simulation models

The ACC-simulation in PELOPS

The traffic simulation program PELOPS (Program for the DEvelopment of Longitudinal micrOscopic traffic Processes in Systemrelevant environment) is orientated to the fundamental elements of the traffic, as there are stretch and environment, driver and vehicle (fig. 5)


Fig. 5: The four main modules of PELOPS

In a modular program structure each of these elements is modelled with a defined interface. The Stretch and environment model allows a detailed description of the influence of a stationary traffic environment like gradients, curves or signs. This can be extended by adding stretch defined parameters like wetness, slippery surface conditions, etc. The traffic element Driver is devided into a decision- and a handling model. In the decision model, parameters of the local driving strategy like the current desired speed and lane are determined. The handling model transforms the characteristics of the local driver strategy into vehicle controls like accelerator pedal or gear lever. Based on these controls, the vehicle model calculates the vehicle dynamic (fig. 6)



Fig. 6: Vehicle model for ACC

The vehicle model is based on the principle of cause and effect, which enables the investigation of autonomous vehicle controls like cruise control, ABS etc. Furthermore the impact of these systems on fuel consumption or emissions of a certain vehicle can be analysed as well as the impact on macroscopic parameters like traffic load, dynamic platoon behaviour etc. In the ACC-simulation the interface between the handling model and the vehicle model is effected. Each time step the ACC-controll algorithm and the driver model calculate the action for the actual situation. If the ACC-system is in process, then the output of the driver model is ignored and the ACC-system regulates the vehicle controlls. The ACC-system can be switched off by the behaviour model, if the desired acceleration or decelleration is outside the operating range of the ACC. Otherwise the ACC is in process. Another feature for the analysis of ACC-systems is the sensor model. It's concept enables the geoemtric description of the sensor beam in different variations (fig. 7)



Fig. 7: Sensor model

The sensor geometry is described through the number of beams, the angle of each beam and the range of each beam. Inside the beam's all vehicles are trated as correct detected. The output of the model is the rate of correct and wrong detections for different sensor geometries. The aim of the geometric sensor design is to detect the relevant vehicle on the actual lane. In the Mitsubishi Diamante (fig. 2), the problem of wrong detections is solved by immage processing. This method is expensive, so that solutions without immage processing can save cost's. For single sensors the detection of wrong vehicles is rising, if the angle and the range of the beam is enlarged. In contradiction to this smaller angle and range cause problems for the longitudinal controll. With the sensor model it is possible to analyse different sensor concepts in a realistic traffic and stretch environment.

Analysis of ACC-systems

This analysis is based on an ACC-system with two actuators, one for the throttel (ac1) and one for the brake (ac2) control. A limit for the braking is fixed, so that the ACC can't cause collisions through high decellerations. The maximum accelleration is also limited for comfortable driving. An actuator for switching the gearbox is not included. The automatic gearbox of the regarded vehicle is switched by the conventional gear change strategy. At first some general aspects of the ACC-system are described. Afterwards the impact on the traffic is outlined.

Effectiveness of the controller concept

Beside the configuration of actuators, the design of the controll algorithm is a very important aspect of an ACC-system. All considerations about the controll strategy must take into account that the driver must accept the whole system. Therefor it is substencial that the actions of the ACC-system are orientated to the driver behaviour. Aspects like safety, driving comfort, and traffic efficency must be taken into account in the controller design. Figure 8 shows the frequency domaine characteristics of a driver and an ACC-controller, which is designed on the above mentioned aspects.


Fig. 8: Controller design

The characteristic of the controller is similar to the driver. The controll error at higher frequencys (above 0,3 rad/s) shows the comfortable concept. Speed changes of the leading vehicle in these frequencyes are damped, so that the speed cycle of the vehicle becomes smoother. A controller concept without this damping would cause alternating accellerations and decellerations in high frequencies which would not be accepted by the driver. The phase delay of the ACC-controller is lower than the drivers phase delay. Above 0,6 [rad/s] the ACC-controller's phase delay is nearly constant at -90°, whereas the driver's is rising. With this characteristic the controller's response is faster than the driver's. This means an improvement of safety. In contradiction to this safety effect, the magnitude shows an ampilifying area in the range between 0,1 [rad/s] and 0,3 [rad/s], which decreases the safety. This means, that the controller is not platoon stable, because speed changes of the leading vehicle are enlarged for each following vehicle. Collisions can be the result of the unstability, if the platoon consists of many vehicles. Figure 9 presents the speed cycle of a 10 vehicle platoon.



Fig. 9: Platoon stability

The platoon is homogeneous, which means that the vehicle's and the ACC system's are equal. All vehicles are in follow mode to the leading vehicle, who's speed cycle (thick line) was measured on a motorway with a medium traffic load. The above mentioned frequency of amplifying has a period between 20 [s] and 60 [s], which appears several times in the cycle. In these periods each following vehicle amplifies the speed changes of the leading vehicle. Between 100 [s] and 150 [s] it can be seen very clear, that the speed course of the last 3 vehicles tears to the lading car's. This platoon characteristic can not be accepted for safety and comfort reasons. There are some characteristics effecting the platoon stability. One possibility to reduce the instability is the enlargement of the desired headway (assumed that the desired headway is depending on the speed). Figure 10 shows the magnitude of equal controllers for a desired headways of 1,2[s], 1,4[s] and 1,8[s].


Fig. 10: Influence of desired headway on platoon stability

At 1,4 [s] the magnitude amplification is smaller than at 1,2 [s], but there are still amplifying frequencys. A further enlargement on 1,8 [s] deletes all amplifying frequencys. Regarding measured headways on motorways it can be gathered that 1,4 [s] is the maximum headway which will be accepted by the drivers, so that in spite of the platoon stability it is not senseful to use such large distances. A possibility to enlarge the desired headway without loosing the acceptance and efficency is a communication based system, where each vehicle is not controlled by the vehicle ahead but by the first vehicle of the platoon. The big disadvatage is, that the system is not autonomous. In this concept, the desired headway rises with the position in the platoon. If the first vehicle has a desired headway of 1,2 [s], then the second has 2,4 + lengthleading vehicle [s] and so forth. Figure 11 shows the speed cycle of such a platoon.



Fig. 11: Improvement of platoon stability

A comparison with figure 9 brings out the effectiveness of the headway enlergment. The platoon has become stable except two situations (80 [s] and 150 [s]) where the second vehicle (which is still unstable with a desired headway of 1,2 [s]) ampifies the speed course of the first vehicle. This can be accepted, because the amplifying is very small at the second position. Another interesting aspect of an ACC- system is the effect on the fuel consumption. Regarding the ACC-concept, which smooths the dynamic of following vehicles, it should reduce the fuel consumption. The following figure 12 shows the operation points in the engine map for the 1. and the 10. vehicle of an homogenous ACC-platoon, which is a stable platoon.



Fig. 12: Fuel consumption

For the 10. vehicle the average value of all operation points is in the area of lower fuel consumption, because the power consumption decreases with the smoothing of the vehicle's speed course. There is also an decrease in the standard deviations of torque and engine speed, which is characteristic for the smoothing. This saving of fuel can be reached only with platoon stable ACC-controllers. If the platoon is not stable, the fuel consumption decreases only for few vehicles behind the platoon leader. For the other vehicles the fuel consumption rises, because of the increse of the power consumption. All descriptions above are based on sensors which detect the distance and speed difference to the relevant vehicle ahead with no error. In reality the sensor is one of the most critical section of an ACC-system. Beside the measurement error the selection of the relevant vehicle can cause system failurs. Basis for the target selection is the sensor geometry, which causes the range of targets. Figure 13 presents the rate of correct detections for a wide and a small sensor geometry.



Fig. 13: Sensor geometry

Basis for the analyzed scenario is a 3 lane motorway with medium traffic load. The wide sensor geometry is a 3 beam sensor with an angle of 9 [°] and a maximum range of 200 [m]. The small sensor geometry has only one beam with an angle of 3 [°] and a maximum range of 100 [m]. It is ovious, that the small sensor has a higher rate of correct detections which is 96,48 [%]. With the wide sensor the rate is only 53,44 [%]. It seems to be useful to keep the geometry as small as possible, but this causes problems in the longitudinal vehicle controll. The difference between both sensor geometrys shows, that a careful design of the geometry can improve the functionality of the ACC-system very much.

Impact on the traffic

The traffic impact of ACC is represented by one analysis which clarifys the effects of the system. This scenario is based on the traffic, shown in figure 14



Fig. 14: Sample szenario

The figure describes the location of the stretch and the measured traffic situation's. From this measurement a time frame of 12 min. is choosen as a scenario, where the traffic load is high. This scenario is modeled in PELOPS as basic scenario. In the ACC scenario's, fourty percent of the vehicles from the basic scenario are equipped with the analysed ACC-concept, to clarify the effects of the system. In three ACC-scenarions the desired headways for the ACC-controll is varied. Figure 15 compares the headway distribution for the basic scenario and the three ACC-scenario's.



Fig. 15: Headway distribution

It is obvious from the basic scenario, that the normal drivers use headway's which are clearly shorter than the required headway of 1,8 [s]. The maximum percentage is in the area between 0,8 [s] and 1,2 [s]. From this stage it is difficult to improve the traffic throughput by an ACC, when the ACC-headways are greater than the driver's. Only the ACC-scenario with a desired headway of 0,8 [s] causes lower values than the basic sceanrio. In the other scenarios (1,2 [s] and 1,8 [s]) the headway values are greater than in the normal traffic. With these larger headway's the traffic throughput decreases, when the speed level remains. From the safety view this headway enlargement is an improvement, because the driver's headway level of 0,8 [s] to 1,2 [s] is approximately equal to the driver's reaction time. Regarding the driving comfort, ACC can be analysed by the accelleration distribution (fig. 16)


Fig. 16: Acceleration distribution

ACC reduces the percentage of high decellerations for all desired headways. The class up to -1 [m/s2] is decreased from 1,6 [%] in the basic scenario to 0,3 [%] in the ACC- scenarios. The area of low accellerations and decellerations (-0,05 [m/s2] to +0,2 [m/s2]) is enlarged. This expresses, that the aim of ACC, to improve the drivers comfort, can be realized. The effect of ACC on the traffic safety can be stated by the 'Time to Collision (TTC)', which is calculated from the speed difference and the distance. Figure 17 compares the Time to Collision distributions.



Fig. 17: Time To Collision distribution

In this diagramm the effectiveness of ACC is very clear. The system reduces the percentage of low Time to Collision, which means an improvement of the safety. Already in the lowest class up to 10 [s] the TTC-reduction can be recognized. This tendency continues in the other lower TTC-classes and causes the TTC increase in the upper classes. The reduction of low TTC's is very imortend, because they are the most critical ones for accidents. This tendency is valid for all headways. The main reason for this effect is the smoothing of the traffic, where the speed differences are reduced, so that the TTC becomes larger. Fianlly the ACC-effect on the road capacity is represented in figure 18 :



Fig. 18: Fundamental diagram

The diagramm consits of the average speed and the traffic flow for the different headways. An capacity improvement means, that speed and traffic flow are higher than the curve without ACC. This applys only to the headway of 0,8 [s]. With a desired headway of 1,2 [s] the situation is close to the basic situation, whereas a 1,8 [s] headway worses the capacity. An explanation for this could be found in the headway distribution (figure 15). It is striking, that only the headway distribution of 0,8 [s] is below the basic headway distribution in figure 15 and that this headway improves the road capacity in figure 18. In both diagramms the situation for 1,2 [s] is close to the basic and for 1,8 [s] the headways are larger than the basis with what the capacity becomes worse.

Conclusion

The traffic simulation program PELOPS enables a wide analysis of ACC-systems. The regarded ACC-concept is based on a throttle and a brake actuator, which are set by a conventional controller. The controll strategy limits the maximum speed, accelleration and decelleration. One of the main aspects in the characteristic of the controll strategy is the relation to the driver behaviour, which requires a smooth controller. Nowadays the ACC-controller have the problem of platoon unstability, which increases the speed course of a leading vehicle. In large platoons this characteristic can cause accidents, because each vehicle amplifies the spead cycle of the vehicle ahead. With an enlargement of the desired headway, the controller stabilizes. If a platoon is stable, then a reduction of the fuel consumption can be achieved with ACC-systems. Beside the problem of platoon unstability it is substencial for the ACC-functionality that the distance sensor detects the relevant target, for example through a special beam geometry, because the rate of correct detections decreases a lot with an unsuitable concept. The main characteristic of the traffic impact is the smoothing of the traffic course. In connection with this, the improvement of the drivers comfort is statet by reduction of high decellerations and accellerations. There is also an improvement of the safety noticed, which is indicated by a clear enlargement of the 'Time to Collision' values. In contrast to this there is no clear improvement of the road capacity. Only very low headways cause an increase in the traffic flow and the average speed. To sum up, ACC is a system which improves the traffic system of today, but there is still a need for further development.