DEVELOPMENT AND IN-VEHICLE APPLICATION OF AN ACC-CONTROLLER

32. ISATA in Wien vom 15.6.99

M. Weilkes, Dipl.-Ing. K. Breuer,
Institut für Kraftfahrwesen Aachen
J. Ludmann,
Forschungsgesellschaft Kraftfahrwesen Aachen

---

Summary
Introduction
The Simulation Tool PELOPS
The Test Vehicle
Simulation of the ACC and the Test Vehicle in PELOPS
Controller Concept
Proof of the Functionality in the Vehicle and Simulation of Traffic Scenarios
Literature

---

Summary

Today's vehicle development does not regard the vehicle as isolated item any more, but understands it instead as part of the whole system human, environment and traffic. For the analysis of the vehicle's influence on this whole system as well as also for an investigation of the interchanges between vehicle, human kind and environment the sub-microscopic traffic flow simulation program PELOPS is developed and applied at the Institut für Kraftfahrwesen.

One of the current research focal points is the development of an ACC controller for the lower speed range and the realization of this controller in the vehicle. The controller is calibrated by means of the simulation of different synthetic and real traffic scenarios in PELOPS.
The following article presents the simulation tool PELOPS and describes the way of the controller development as well as the first application results of a simple controller in a test vehicle.

Introduction

In recent years, the rapid development of electronic and communication technology has led to the investigation and development of new kinds of vehicle and traffic systems. One group of these systems, the driver's warning and information systems, could contribute substantially to an improvement of driving comfort and safety in the future. As a system within this group the 'intelligent Cruise Control? (Adaptive Cruise Control, ACC) was developed. An ACC-system determines the distance and differential speed in relation to the preceding vehicle by means of a distance sensor and then calculates by means of the own speed a safe distance as time gap to the preceding vehicle. This distance is controlled by electronic intervention into the acceleration and braking pedal. The system works vehicle autonomously, meaning that the vehicle collects by sensors all necessary data on it's own for the vehicle control.

The first generation of ACC-systems was designed for the higher speed range and thereby for an application on highways and well-structured country roads and is currently going into mass production. The first step of the ACC-development thereby being concluded, the current research activities focus on the functional extension of the system for an employment in city and suburban areas.

This extension of the ACC functionality seems to be making sense also because of the often recognizable tendencies in city traffic planning to separate efficient and enlarged traffic roads from traffic abated residential districts. By means of this separation, apartment areas worth living in are created on the one hand. On the other hand, high traffic performance without obstructions on through roads has to be achieved on a relatively high velocity level, in order to use the remaining traffic area as efficient as possible. Here it is necessary to avoid the economically and ecologically negative effects of traffic jams. At the same time means have to be found to reduce the rising driver's stress caused by traffic growth and complex traffic situations and the thereby resulting accident potential.

However, the development of assisting systems for drivers warning and information, such as ACC, is highly complex and has to take into account all system components, starting from data registration by suitable sensors over the evaluation and processing of the measuring data up to the consideration of the driver. Apart from the optimization of the system components it has to be investigated at the same time, how the application of assisting system affects the neighboring vehicles or the overall traffic. This overall approach requires new development and analysis tools. Such a tool must be able to reproduce the vehicle as well as the assisting system in detail and simultaneously be able to depict the surrounding environment (meaning driver behavior, stretch topography, traffic signs etc.). With the program system PELOPS such a tool is available. The following paper presents the functionality of PELOPS by the exemplary means of the development of an ACC-controller for the lower speed range.

The Simulation Tool PELOPS

PELOPS was developed at the Institut für Kraftfahrwesen in co-operation with BMW. The idea of PELOPS lies in the combination of detailed sub-microscopic vehicle models with microscopic traffic models that enable an investigation of the longitudinal vehicle behavior as well as an analysis of the traffic course. The realization of such a complex tool would never have been possible without the rapidly increasing calculation power of computers.
In it's structure PELOPS is orientated at the most important elements of the traffic system - stretch/environment, driver and vehicle. In a modular program structure the named elements are modeled and separated by interfaces (figure 1).



Fig. 1: PELOPS-Structure

The stretch model allows, if necessary, a detailed description of the influences of a stationary traffic environment. The course of the road in horizontal and vertical direction is indicated by radius and transitions as well as the number and width of the lanes. In addition to these geometric data also traffic signs and environmental conditions by parameters such as wetness, slipperiness etc. can be given in advance. The actual traffic conditions for a vehicle result from the number of surrounding vehicles and their distances and speeds. In order to be able to depict certain ways of behavior in traffic or in order to be able to drive given driving cycles again and again, also certain speed profiles can be set for the driver-vehicle-units.

The vehicle model bases on the 'Cause-and-Effect-Principle' [1]. It calculates the driving force starting from the engine operation point over the clutch, transmission and differential up to the wheels, where the driving force is then balanced with the driving resistances. The operation point is changed by the alteration of the motor torque (cause). From the thereby caused acceleration and speed change results the engine speed (effect) under consideration of the drive-line elements. As transmission modes the conventional hand shifting as well as automatic gear are implemented. For commercial vehicles a retarder can additionally be depicted. Only such a detailed description of the vehicle under employment of the cause-and-effect-principle allows the investigation of control engineering equipment such as ABS and ACC.

The link between vehicle and traffic simulation is represented by the driver model. It is divided into a behavior and a handling model. In the behavior model the parameters of the local driving strategy is determined by means of the actual driving situation and the vehicle environment. The parameters of the local driving strategy are the desired acceleration, lane and eventually also the shifted gear. In the handling model these parameters are then converted into vehicle-specific quantities such as acceleration pedal, brake etc.
For the determination of the driver behavior PELOPS works with a psycho-physic follow-the-leader model that bases on the works of Wiedemann [2]. For the application in PELOPS this model was adjusted and significantly further developed /3,4/. Because the follow-the-leader model takes only the reaction to vehicles in the same lane into account, additionally, a tactic behavior was developed that depicts a realistic driver behavior in relation to the stretch's course and in case of multilane stretches. This tactical behavior includes for example the reaction to slow vehicles on the neighboring lane, to different traffic signs and intersection forms and a situation-depending lane changing model.

The Test Vehicle

As basis for the test vehicle a regular BMW 325iA (E36) is used being ex factory equipped with an automatic gear shift and a conventional Cruise Control. In order to fulfill the requirements of the special applications the vehicle was additionally equipped in comparison to the series vehicle.

The detection of the environment necessary for the ACC functionality is being done by two sensors integrated into the vehicle's front. They are a laser sensor by Leica and a pre-series model of the radar sensor being installed in the Mercedes S-class (W220) as of spring 1999. Furthermore, the test vehicle disposes of additional sensors for the detection of the driving state, for example acceleration, gear rate and speed sensors. As operation elements for the driver the states of acceleration- and braking pedal, direction indicator- and Cruise control lever as well as the gear lever are recorded.

The possibility to intervene into the braking, the throttle and the gear choice of the vehicle is given. In order to do so, an electronically controllable vacuum-brake booster substitutes the series component. It can reduce the speed with a deceleration up to 6.1m/s2, if necessary, even without the intervention of the driver. Furthermore, the position unit of the throttle being used by the regular Cruise Control is also used by the ACC-system. By interventions into the transmission control the second, third and fourth gear can be directly selected. An adjusted electrical connection ensures that in normal driving situations without ACC the standard Cruise Control, the normal brake and the manual gearshift choice are always available.

A central control unit on the basis of a 16-bit Siemens 80C167 micro-controller processes the recorded measuring data and controls the actuators. The communication between the single components of the ACC-system is being done by a CAN-Bus that was subsequently being installed in the vehicle. System components that do not dispose of an own CAN-interface are connected to the bus by means of SLIO-modules that base on an 8-bit Philips 80C592 micro-controller.

The ACC-operation program supplies the controller with all recorded data and expects a desired brake pressure and a desired throttle position as input data that are converted by a state regulation.

Several hardware- and software supported fail-safe-levels ensure during the ACC operation that eventually occurring errors of a system component or of the control unit are detected, the driver is warned and that the ACC is deactivated.

Simulation of the ACC and the Test Vehicle in PELOPS

The aim for the development of the controller in PELOPS was to provide a depiction of the environment as close to reality as possible. This realistic reproduction on the one hand refers to the vehicle itself and on the other hand to the data recorded by the sensor and the interaction between driver and system.
At first, the vehicle was depicted in PELOPS with the vehicle data and the characteristics of the engine, the hydrodynamic converter and the gear switch- and lockup characteristic lines. For a first comparison between simulation and reality a calibration of the simulation was made by having the vehicle drive single cycles on a dynamic roller test bench that were then duplicated in the simulation.
For a simple transfer between simulation and real vehicle an interface was created that is being filled out in the vehicle as well as in the simulation. In the simulation the available data of the vehicle is provided in a corresponding quality and accuracy. Because the control unit in the vehicle should only work with fixed point-calculation, also a conversion of the quantities takes place in the interface. Input data of the controller consists of the sensed current state of the own vehicle and the data of the preceding vehicle registered by the distance sensor. The output data of the controller is a desired position of the actuators, meaning a desired pressure in the braking system or a throttle position.
For the survey of the environmental data a geometric model was used that corresponds to the detection range of the employed radar sensor. In order not to reproduce an idealized picture of the environment the sensor data as well as the data of the own vehicle were overlapped with noises. The noise dimension of the single signals was set depending on a measurement in the vehicle. However, since a middled frequency-free white noise was used in the simulation, the stochastic dispersion of the data is greater in the simulation than in case of real sensors.
For the investigations of the control algorithm in complex traffic scenarios the depiction of the interaction between driver and system is necessary. This boarder for the driver to take over control bases on the empirical investigations after [5] and is determined in dependence on distance, differential speed, maximum braking of the ACC-controller as well as the speed of the preceding vehicle as desired speed. In case of falling below a certain distance



the driver takes over and switches the controller off. Additionally, the driver takes over in front of a traffic light or in vehicle still stand. For the restarting after still stand the ACC-system has to be activated by the driver. Thereby, the driver's reaction time at restarting remains the same even in case of using the system.

Figure 2 shows the inclusion of the ACC controller in the PELOPS simulation environment.




Fig. 2: Inclusion of the ACC controller in PELOPS

Controller Concept

For proofing the transferability of the controller development from the simulation into the vehicle a very simple control algorithm is used at first. The basis of this algorithm is the so-called 'ika-controller' that was already employed in the simulation during the investigation of the column stability in the frame of PROMETHEUS and that is described in [3]. In dependence on the own speed of the vehicle and the time lap set by the driver, at first a target distance dx_target is calculated. By the comparison of dx_target with the measured dx_actual a target different speed dv_target is determined in relation to the preceding vehicle. The target acceleration a_target of the own vehicle is then determined by the deviation of dv_target and dv_actual.

The target acceleration of the controller is converted by the acceleration controller in a throttle position or a braking pressure. Because the hydrodynamic converter causes a non linear behavior in the lower speed range in case of automatic vehicles, the acceleration controller was realized by means of inversion of the converter and engine characteristic lines. In dependence on gear, speed and engine speed the needed engine torque and the corresponding throttle position is calculated for the required acceleration. In order to manage with a low computer power in the vehicle the engine characteristic field was reduced to three characteristic line, in between those an interpolation takes place. An overlying PI correction is in charge of the adjustment of changeable driving resistances and outer obstruction quantities.

Proof of the Functionality in the Vehicle and Simulation of Traffic Scenarios

The acceleration controller as well as the above lying ACC-controller were, at first, designed in the simulation and then separately realised in the vehicle. By means of simulation the reactions of controller parameterisations to different synthetic and real driving manoeuvres can thereby be easily investigated. Fig. 3 shows as an example the reactions in approaching the preceding vehicle for a harder controller variant, which is focused more on target distance, and a softer one, which is more comfort-orientated. A distinct difference in the reaction?s time beginning can be detected that also causes different acceleration and distance characteristics.



Fig. 3: Speed and distance course of an approach for different controller designs

On the basis of such pre-investigations a controller parameterisation was chosen and implemented into the test vehicle. In Fig. 4 the speed courses of a follow-up drive from a measuring and simulation are shown as proof of the transferability between simulation and reality.



Fig. 4: Comparison of measured and simulated follow-up behaviour

In order to enable this comparison, a follow-up drive was executed in automatic mode, in which the preceding vehicle generates a dynamic speed course. Then, the speed course of the preceding vehicle was afterwards reproduced and transferred to the simulation by the data retrieved from measuring technique and distance sensors of the ACC-vehicle. It was set as the speed profile of a vehicle that is also in the simulation being followed by the ACC-vehicle with the same controller algorithm. After the initialisation of the simulation with measured starting conditions it is then run freely. The thus gained simulation data are compared to the actually measured data in Fig. 4.

Apart from the possibility of the controller design at simple synthetic and actually measured speed cycles the simulation with PELOPS offers the possibility to analyse the effects of different controller design on the traffic flow in complex traffic scenarios. As example a scenario from inner-city traffic serves here, as it was registered with comprehensive measurements in [3] and was validated in the simulation. In this scenario two different controller versions were investigated. One version is designed for column stability, whereas the other is more comfort-orientated and generates softer transitions of speed.

In Fig. 5 the distinctly varying effects of the two controller versions on the traffic flow are presented in case of different ACC-equipment levels. The changes in percentage to the basis simulation, which reproduces the reality, caused by the different equipment levels are shown. The considered characteristics are the travel-time, the number of driver overtaking and the average speed at a local measuring point.



Fig. 5: Influences of different controller designs on the traffic flow

The example indicated the great influence of changes in controller parameters on the traffic. Through the integration of the possibility for system design and investigation in complex traffic situations the opportunity is given by PELOPS to register the interchanges at a very early stage.

Literature

[1] DAVID, Wolfgang
Modulares Simulationsprogramm zur Auslegung und längsdynamischen Untersuchung von Kfz-Antriebssystemen (Modular Simulation Program for the Design and Longitudinal Investigation of Vehicle Drive Systems), Ph.D. Thesis at the Institut für Kraftfahrwesen, Aachen, 1992

[2] WIEDEMANN, R.
Simulation des Straßenverkehrsflusses (Simulation of Traffic Flow), Schriftenreihe des Institutes für Verkehrswesen der Universität Karlsruhe, Heft 8, Karlsruhe, 1974

[3] DIEKAMP, Rainer
Entwicklung eines fahrzeugorientierten Verkehrssimulationsprogramms (Development of a Driver-orientated Traffic Simulation Program), Ph.D. Thesis at the Institut für Kraftfahrwesen,
Aachen, 1995

[4] LUDMANN, Jens
Beeinflussung des Verkehrsablaufs auf Straßen - Analyse mit dem fahrzeug­orientierten Verkehrssimulationsprogramm PELOPS (Influences on the Traffic Flow on Roads - Analysis by means of the Traffic Simulation Program PELOPS), Ph.D. Thesis at the Institut für Kraftfahrwesen, Aachen, 1998

[5] KOPF, M.; NIRSCHL, G.
Driver-Vehicle Interaction while Driving with ACC in Borderline Situations. Proceedings of the 4th World Congress on Intelligent Transport Systems, Berlin, 1997.