Active Steering Systems are evolving in the direction of giving an increasing support in terms of agility, safety and comfort improvement.
The automotive industry is showing a growing interest towards the usage of EPS (Electric Power Steering) systems. With the enhancements in mechatronic technologies, they are spreading from small size cars to medium and large size cars. EPS systems provide a number of advantages with respect to hydraulic power steering: first of all lower fuel consumption, then higher product standardisation and finally easier customization of feeling performances. Nowadays, EPS systems are extending their influence from the driving feeling optimisation to other driving features including safety, agility and comfort. This is achieved by modifying the conventional assistance torque, based on driver torque and steer angle, through the introduction of contributions based on signals that other ECUs, integrated in braking, vision, obstacle detection systems, compute and make available to the vehicle CAN network. These “torque overlays” determine a useful haptic feedback for the driver, recommending specific wheel steer commands suitable for the peculiar dynamic scenario of the vehicle.
Hereafter, it is shown that using the available methods of system theory and control design, it is possible to see both the fundamental problems of feeling improvement and the design of torque overlay contributions in the same perspective, only modifying the system to be studied and reformulating properly the target to be pursued. This methodology has been successfully tested for the case of steering feeling improvement during a test campaign at FIAT test track in Balocco (IT) and is going to be validated for the design of a torque overlay system dedicated to lateral dynamics control enhancement.
Usually, when a new vehicle model has to be launched in the market, the steering engineer has to not only design the EPS control from scratch but also consider an existing version on a similar vehicle and to adapt it to the brand and customer expectations. In this frame, the problem of feeling enhancement answers to this basic question: how to shape existing assistance in order to modify feeling performances accordingly to desired targets?
First of all, this question implies a methodology to associate specific SW modifications to deterministic improvements of the steering feeling. Second, it implies the capability to transform the feeling, something that is by definition subjective and not univocal, in something as much as possible objective and measurable. Trying to answer that question for the case of existing Fiat Stilo EPS improvement, in view of the new Bravo model, in the course of 2005, a team with members from Fiat and its research centres (CRF-Elasis) has been involved in the activity of EPS enhancement hereafter described in order to:
The work methodology, well sketched by the following picture, is represented by two stages: the study phase, based on a steer-EPS-SW simulator tuned on the track, that proposes control modifications to achieve the previous listed aims, and the development phase in which the validation/rejection of the proposed modifications is accomplished on the track using a vehicle provided with an RCP system.
As far as the methods for objective classification of results are concerned, two metrics have been chosen: the SMA (Sweep Manoeuvre Analysis) and the IQS method (Steering Quality Index).
Figure 1
The SMA is based on a direct interpretation of the steer behaviour in the frequency domain, measured with a steer sweep manoeuvre at constant speed. In particular, the SMA monitors gain and delay in the range between 0.5 and 1Hz, and the peak of maximum gain in terms of resonance frequency and ratio with respect to steady-state gain. Through previous experiences on both electrical and hydraulic power steering systems, a correlation of these variables with specific steering feelings has been found, and performance targets have been set. Those experiences showed that a gain curve as “flat” as possible is preferable, and that an eventual negative delay or an excess of positive delay must be avoided. The IQS synthesizes the steer performances by using objective parameters measured (computed) in a number of significant manoeuvres: 0.2 Hz sinusoids; step steer and release with several steer angles at 100 kph; parking cycles. The procedure identifies a set of eight sub-indices relevant to deadband, graduality, elasticity, centre feel, behaviour during steer release, effort in parking, and a global quality index of the steering (IQS).
These two evaluation metrics may be performed on data obtained through either experiments or simulations; therefore they can be also exploited to verify the steer-EPS_SW model suitability in representing the steer behaviour of the true EPS mounted on the true vehicle on the track.
The system analysis of the model permits to write the complete transfer function Dvol/Cvol associated to a sweep manoeuvre in dependence of elementary transfer functions TF related to the mechanical steer system, to the EPS algorithms and to the road-vehicle interaction. If everything has been correctly described and tuned in the frequency domain, the Dvol/Cvol transfer function, computed from this formula, must be more or less coincident with the one coming from the same sweep trial on the track. Exploiting this bi-univocal relationship between Dvol/Cvol and assistance, a reciprocal formulation allows to determine which adjunctive contribution, for instance in the form of a shaping filter, could be integrated in the SW to fulfil a given desired modification of the Dvol/Cvol characteristic.
Figure 2
Once these hypotheses of modification have been determined, the metrics above defined can help to verify, by simulation, if they are promising or not. The final point before going to the track is the verification of the modified EPS control in terms of robustness assessed by mu-analysis tools [2] versus possible structured uncertainty for instance, due to the tyre or to the EPS motor.
Figure 3
The evaluation of the proposed EPS control/tuning improvements in the vehicle have been performed with the use of an “Open-EPS” system. The Open-EPS is mechanically identical to NP-EPS (manufactured by TRW), and its software has been modified in order to connect it, via a dedicated CAN network, to an RCP system from dSPACE, where an easily modifiable copy of the EPS high-level software is running.
The assistance torque effectively actuated may be alternatively the one calculated in the RCP or the one calculated by the original high-level SW in the EPS; the switching between the two modes is implemented in an immediate and easy way, through a hardware switch connected to the RCP. This capability has been demonstrated to be fundamental for the subjective trials in fact, testing consecutively the original and modified algorithms in the same manoeuvre, made easier the evaluation of the qualitative differences.
Figure 4
The two pictures hereafter attached show a synthetic view of the experimental results derived from objective test sessions. Note that in addition to initial and final versions, plots report an intermediate SW version obtained by modifying parameters and existing logics without introduction of described assist shaping. The diagrams relevant to the sweep manoeuvres show that the gain curve has become flatter and that the delay has become neither negative nor excessively positive. Different amplitudes steer pulses executed at a speed of 100 kph have permitted also the verification of enhancement damping features. Subjective evaluations from professional drivers have confirmed the good SMA and IQS rating of the modified control.
Objective of this second step is to explore torque overlay strategies for an integration of active steering and active braking functions devoted to:
Now the basic question becomes: how to define torque overlay functions that are able to modify the dynamic performances of the vehicle accordingly to desired targets? It can be seen that instruments and results achieved from previous step may help in this direction.
Figure 5
First of all, it is necessary to enlarge the boundary of the previous “feeling problem” (red dashed line in the below picture), by including a driver model and a lateral dynamics vehicle model.
Then, taking into account that the torque overlay has the final aim to recommend virtuous corrections of the steer angle in specific critical dynamic situations, the transfer function Vy/VyREF is chosen as suitable target to be improved. Objectives of the work in this case corresponds to augment rejection on the peak at resonance frequency and to reduce the phase delay in the range up to about 3Hz where the cooperation between a system like this and the driver is still meaningful. It is immediate to see that the first objective corresponds to lower overshoot and the second one corresponds to larger promptness (faster convergence) to the steady state conditions. Due to the presence of the driver in the loop, it is worth to underline the large range of possible vehicle dynamics behaviours. As a general rule, the system is well designed if, for the same driving styles, those behaviours are improved by the presence of the torque overlay.
Figure 6
Analogously to the previous “feeling problem” it is possible to determine the Vy/VyREF transfer function by exploiting the knowledge of the elementary transfer functions relevant to vehicle lateral dynamics (an identified single track model is sufficient at this level), steer system and conventional EPS-SW (the ones already discussed for problem in section 2) and driver (an Hess model has been adopted). Once this step has been accomplished it is possible to reformulate the generic basic question before in the following manner: how to find functions, based on specific lateral dynamic measurements that determine a steering assistance modification so getting the desired improvements of smoother and faster dynamic behaviour? For the case under study, two feedbacks have been built based on yaw rate and, respectively, lateral velocity derivative error that, in stationary conditions, is proportional to the side slip rate. This information is directly available on the vehicle using the gyro and the lateral accelerometer and this makes easier the implementation. In order to fulfil the prescribed objectives, the two feedbacks have been computed and proved in a Matlab-Simulink® simulator demonstrating that:
Active Steering Systems are evolving in the direction of interaction with several vehicle features: agility, safety and comfort. From the implementation viewpoint, some of these improvements can be obtained providing a haptic feedback to the driver through a torque overlay superimposed to the conventional assist torque. Using the instruments of system theory and frequency analysis it is possible to transform requirements, described in words, in specific modification requirements for peculiar target transfer functions. Exploiting knowledge of the various elementary transfer function (steer system, EPS-SW, road-vehicle interaction, vehicle dynamics and driver), it is possible to determine the a-priori behaviour of the target transfer function and, by comparison with its desired behaviour, modifications can be proposed to the existing EPS logics.
Keywords: Electric Power Steering, Active Steering Systems, Park Assist, Lane Keeping Departure Warning, Pull Disturbance Compensation, Fiat Stilo, Steer Behaviour, Dynamic Torque Overlay
