Principal heading

DSC 2007 North America – Iowa City – September 2007 Introduction
It is still difficult to simulate vehicle motion in a realistic way. Incorrect simulator motion even causes simulator sickness (Kennedy, R.S. et al., 1992). Therefore, the use of motion in simulators is still subject to debate. In order to explore the requirements for proper motion simulation, an advanced motion platform is required. Desdemona provides just res enhanced performance and new capabilities as compared to standard hexapod designs. Desdemona is a moving-base research-simulator located at TNO (Soesterberg, The Netherlands) that was designed with a special focus on spatial disorientation demonstrations, flight simulation, and driving simulation. It was built in close co-operation with AMST (Ranshofen, Austria). The simulator has 6 Degrees of Freedom (DoF) and is in principal a centrifuge design with additional DoFs (). The cabin is mounted in a gimbaled system (3 DoF, >2π radians), which as a whole can move vertically along a heave axis (1 DoF, ±1m) and horizontally along a linear arm (1 DoF, ±4m). This structure can as a whole rotate around a central axis to facilitate centrifugal motion (1 DoF, <3G). The six degrees-of-freedom result in a large cylindrical motion space with a wide dynamic range as compared to conventional hexapod design. Figure 1: Desdemona advanced motion platform
The design offers new possibilities with respect to motion cueing in relation to human motion perception, in particular for driving simulation (Wentink et al., 2005). For example, to enable sustained acceleration, the system can centrifuge around the central axis while the cabin is at an eccentric position somewhere on the linear arm. Within the European Eureka project ‘MOVES’, Desdemona (amongst other simulators from Renault, DLR and Max Planck Institute) is utilized to investigate to which extent motion cueing in driving simulation can be optimized. The research aims at determining how realistic driving with such an innovative motion-platform can be and what the limitations are. __________________________________________________________________________________________ DSC 2007 North America – Iowa City – September 2007 Figure 2: The 6 DoF of the Desdemona moving base simulator
This paper describes the Desdemona simulator in a car driving application. The setup of the different components required for car driving is briefly described first. The next section discusses an example of recorded real-world vehicle data. Thirdly, this curve-driving data is used to demonstrate basic, but innovative, motion filter design principles for driving simulation in Desdemona. Finally, conclusions are presented. Desdemona in car driving simulation
This section describes briefly the setup that is used for car driving. The cabin of Desdemona can be equipped with a car mock-up that contains force-feedback on pedals and steering wheel and out-the-window visuals with 120x40 degrees visual angle. Direct drive electrical motors generate the control loading for the steering wheel, gas and braking pedals. Electrical motors - compared to springs - enable a flexible way to modify the loading characteristics or to test and evaluate driver-support systems such as a haptic gas pedal (Rook and Hogema, 2005). The characteristics of the pedals are programmed in an I/O computer inside the cabin in order to achieve a high bandwidth. The vehicle model - located at the ‘shore’ – calculates the input for the steering torque. The sample frequency of the pedals is 1000 Hz and the sample frequency of the vehicle model is 180 Hz. The car mock-up communicates via shared memory to a Matlab-Simulink environment in which the vehicle model is running (Hogema, J. et al., 2004). The vehicle model is an s-function generated in Carsim (Ann Arbor, MI, USA). The motion cueing filter runs in a separate Matlab-Simulink thread, again via shared memory. The motion cueing filters receive the six DoF from the vehicle plus the first second order time derivative. The __________________________________________________________________________________________ DSC 2007 North America – Iowa City – September 2007 output of the motion filter is send via a ‘bridge’ computer to the Desdemona control computer. The PC based Computer Generated Image system consists of three computers that generate real-time images at an update-rate of 60 Hz. The sound generator is a sampled sound system, simulating the sound of the own vehicle (tyres, engine, wind) as well as other traffic. Collection of sample car data
In order to develop motion cueing algorithms for car driving, real data were collected of some typical car driving situations in an urban environment. The experiment was conducted using TNO’s instrumented vehicle. This is a Volkswagen Passat with automatic gearshift. An Inertial Measurement Unit (IMU) (MMQ 50, Systron Donner Inertial Division) was placed in the plane of symmetry of the car, between the two front seats, inside the compartment covered by the armrest. The IMU measured the three linear accelerations and the three angular speeds. The measured signals were anti-causal smoothed by a second order Butterworth filter. The measurements were used to roughly estimate the shape and size of city curve driving, therefore the data were not compensated for drift and the positioning of the IMU with respect to the driver’s head This paper elaborates on one driving maneuver, namely curve driving. The dashed line in easured angular speed when driving a curve to the left with a total angle of 90 degrees. The y-axis is positive to the left with respect to the driver (lateral), the x-axis is pointing to the front of the car (longitudinal), and the z-axis is positive up. The dashed line at the bottom plot shows the lateral specific force. The lateral specific force is positive because a turn to the left was taken. Figure 3: Measurements (two) and angular approximation of the angular speed and the later specific
force when driving a curve of 90 degrees.
__________________________________________________________________________________________ DSC 2007 North America – Iowa City – September 2007 In order to generalize the motion filter design, the measured signals were approximated by an analytical solution as indicated by the solid). A cycloid approximates the heading, and Equation 1 describes the associated angular rate during the curve. The longitudinal car velocity is V = 4.5 m/s , the total curve angle is A = π / 2 , the =10 m and t represents time. A second order low-pass Butterworth filter (cut-off frequency of 5 rad/s) smoothes the analytical solution to overcome discontinuities at the transition from a straight road to a curve and vice versa. The corresponding specific lateral force f is found by where R is the rotation matrix from the car frame-of-reference to the world frame-of-reference. For the sake of simplicity, we assume zero specific forces in longitudinal and normal direction and zero angular rates in roll and pitch of the vehicle. Motion cueing filter principle
Usually, for conventional hexapod simulators only specific forces are used as input signal for motion cueing filters, i.e. the rotational speeds and other dynamic parameters are not used. In this paragraph, two first-guess filters illustrate the explicit incorporation of the heading rate of a vehicle. In a future experiment the effects of this explicit use of the heading rate needs to be investigated. The previous analytical solution of curve driving is used as input signal. One-to-one copy of the heading rate
Desdemona enables the possibility to couple the heading of a car directly to its central yaw (centrifugal rotation), i.e. a one-to-one copy of the heading. The driver’s nose is oriented perpendicular to the main armconsists of straight and curved segments. When driving a straight road the simulator central yaw rate is chosen to be zero in order to prevent Coriolis forces that generate a false perception of driving. At the transition from a straight segment to a curved segment and vice versa an additional longitudinal force from central yaw onset acceleration is generated in the simulator that was not present in the real car. A second difference with real car driving is based on the smaller radius of the main arm. At a straight section, the curve radius is infinite and at a curved segment it smoothly decreases from infinity to a minimum value and back to infinity again. Due to this, the generated lateral force in Desdemona – without additional measures – will be smaller than in the real world if a one-to-one copy __________________________________________________________________________________________ DSC 2007 North America – Iowa City – September 2007 shows the results without any compensation. The sinusoid is the longitudinal force. Figure 4: A one-to-one copy of the car heading to the central yaw of Desdemona without any
compensation. The figure shows an additional longitudinal force fx and lower lateral force fy.
Compensation by tilting the cabin
The additional longitudinal force and the reduced lateral force can be compensated by using cabin tilt with respect to the gravity vector. As mentioned, due to the acceleration of the central yaw a tangential force is generated pointing in the direction of the driver’s nose (from the driver’s perspective the tangential force has the same direction as the longitudinal force). Consequently, the direction of the specific-force vector does not coincide with the real-world maneuver. By tilting the cabin dynamically in pitch, the specific-force vector can be aligned in such a way that the longitudinal force reduces. The same approach is used to increase the lateral force on the driver in Desdemona by The closed-loop tracking of the real specific-force vector, by changing both the pitch and the roll gimbals is achieved by applying a multi-variable closed loop controller (Wentink et al., 2005). Due to the tilt compensation, the angular rates in the cabin frame change significantly. The magnitude of the rates along the x and y-axis are above the threshold value of 3 degrees per second [Groen, Hosman, TNO-NLR studie]. An angular rotation above the threshold means that the driver perceives these false rotations. The angular rate could be decreased by compensating less strictly the deviations in the specific-force. __________________________________________________________________________________________ DSC 2007 North America – Iowa City – September 2007 Figure 5: The longitudinal and later force is compensated by aligning the specific-force vector with
the real maneuver by using the pitch and roll gimbals (left). The angular rate of the driver is shown
in the right plot (ωx is the roll rate and ωy is the pitch rate).
Compensation by Coriolis force
Another possibility is to partly compensate the longitudinal force by utilizing the Coriolis force. A Coriolis force occurs when the length of the main arm (radius) is changed during centrifugal rotation. The force points in the same direction as the tangential force. Equation es the longitudinal force in the cabin when the gimbals angles have zero speed and orientation. f = − ψ&& − ψ& & The position of the cabin on the main arm is denoted by R and the central yaw angle by ψ (centrifugal). As a side effect, due to the linear acceleration over the main arm and the decrease in radius, the magnitude of the lateral force increases at the start and the end of the ma In order to prevent large roll-tilt compensation for the decrease in lateral force, the Coriolis force is only partly used. The remaining compensation is again done by the pitch __________________________________________________________________________________________ DSC 2007 North America – Iowa City – September 2007 Figure 6: The longitudinal force is compensated by moving the linear arm and by aligning the
specific-force vector with the real maneuver by using the pitch. The lateral force is compensated by
the roll angle (left). The angular rates in the cabin is decreased (right).
The required radius (order Taylor approximation of the natural logarithm of the real-car heading at 0.25. The logarithm of the heading rate was found analytical. Figure 7: The radius of the Desdemona cabin during the simulation.
The difference between is found in the details. First of all, the longitudinal specific force component reduces due to the linear-arm motion (Coriolis). However, some lobs at the start and the end of the maneuver appear in the specific force in y-direction as a side effect (without gimbals motions). The remaining longitudinal force and the deviation in the lateral force are compensated again by roll and pitch tilt. Conclusions
This paper has described the Desdemona simulator in a car driving application. To this end, the different components required for car driving have been briefly discussed and an example of recorded real-world vehicle data has been given. This curve-driving data has been used to demonstrate basic, but innovative, motion filter design principles for driving __________________________________________________________________________________________ DSC 2007 North America – Iowa City – September 2007 simulation in Desdemona. Usually, for conventional hexapod simulators only specific forces are used as input signal for motion cueing filters. In this paper, two first-guess filters have illustrated the explicit incorporation of the heading rate of a vehicle. The first filter used a fixed linear arm radius and compensation by using cabin tilt with respect to the gravity vector. The second filter used Coriolis force and cabin tilt for compensation. The Coriolis force helped for compensating in the longitudinal direction and appeared in the lateral direction as a side effect. In a future experiment the effects of this explicit use of the heading rate needs to be investigated. References
Hogema, J.H., Hoekstra, W., & Stel, I. (2004). On-line vehicle model coupling Driving Simulator - ADVANCE (TNO Memorandum TNO DV3 2004-M 059). Soesterberg: TNO Defence, Security and Safety. Kennedy, R.S., Fowlkes, J.E., Berbaum, K.S. & Lilienthal, M.G. (1992) Use of a motion sickness history questionnaire for prediction of simulator sickness. Aviation Space and Environmental Medicine 63:588-593. Rook, A. & Hogema, J.H. (2005). Effects of Human Machine Interface Design for Intelligent speed adaptation on driving behaviour and acceptance (TNO Report TNO-DV3 2005 D014). Soesterberg, Netherlands: TNO Defence, Security and Safety BU Human Factors. M. Wentink, W. Bles, R.J.A.W. Hosman, &, M. Mayrhofer. (2005) Design & evaluation of spherical washout algorithm for Desdemona simulator. In: AIAA conference proceedings, simulation technologies. __________________________________________________________________________________________

Source: http://www.amst.at/sites/products/desdemona/desdemona-reports/Desdemona_paper_DSC_2007.pdf