Using our knowledge on motion perception, we aim to improve the quality and realism of motion simulation by developing perception-based motion cueing algorithms. A motion cueing algorithm (MCA) converts the motion of a vehicle (or vehicle model) into simulator input commands. What differentiates our algorithms from many existing MCAs is that they incorporate models of human self-motion perception. With these models we can compute the difference between the desired (vehicle) percept (i.e. the motion perception one would have in the actual vehicle) and the provided (simulator) percept. By minimizing the difference between these two percepts, a perception-based MCA can improve the quality of simulation.
In dynamic vehicle simulation, motion cueing algorithms (MCAs) aim to map the vehicle motion that is to be simulated onto the limited capabilities of simulators, while at the same time preserving the perceptual realism of the simulation. The goal of MCAs is therefore to transform the linear and angular accelerations of the simulated vehicle into translations and rotations of the motion platform, such that perceptually equivalent specific forces and rotations are provided to the simulator's occupant.
Most MCAs are based on washout
filters. One feature of a washout filter is that the algorithm always returns the simulator to its initial position over time (in other words, the motion is 'washed out'). A second feature of the washout approach is the representation of translational motions by rotations. Through the filters, the inputs are split into high-frequency and low-frequency components. The high-frequency components are integrated to produce the translational motion of the platform, while the low-frequency components are reproduced by tilting the platform. This technique is called the tilt-coordination
, and is currently one of the most used "perceptual tricks" in motion cueing. Tilt-coordination relies on the tilt-translation ambiguity. This ambiguity arises because under certain conditions, rotational motions (also known as tilt) are indistinguishable from translational motions, especially when a translational motion is concurrently presented visually. For example, by tilting (pitching) the simulator upwards, a sensation of forward acceleration can be achieved. This sensation is produced by reorienting the body with respect to gravity such that the sensory organs are stimulated in a similar way as during translational acceleration. This sensation is enforced by the body being pressed into the seat.
The washout approach has certain limitations. For example, the tilt-coordination technique only works if the tilt rates are limited, such that the tilt itself is not perceived. Otherwise the driver feels the rotational motion instead of the desired illusion of translational motion. The washout behavior can degrade realism too, as stronger motions cause a stronger washout, which may be detectable. Washout filters typically require extensive tuning, mostly done by experts, and the optimal tuning depends on many different factors such as vehicle dynamics, simulator specifications and simulated maneuver.
At the Max Planck Institute for Biological Cybernetics we are developing a new type of MCA that exploits the available knowledge on human motion perception to overcome the drawbacks and limitations of conventional MCAs. This algorithm includes a scalable version of a self-motion perception model (see Self-motion perception models
), and uses its output to minimize the mismatch between desired vehicle motion and actual motion perception in the simulator. Through using perception information, several advances can be made in simulating acceleration profiles in a motion simulator. For example, given the simulator specifications, it can be determined how and how well a simulator can reproduce a motion profile. It can also be investigated how a classical washout filter should be tuned in order to minimize the problems that are inherent to this approach. Finally, using the perception-based MCA that we are currently developing, the mismatch between desired vehicle motion and actual motion perception in the simulator can be minimized.
We perform experiments to validate the perception-based MCAs using psychophysical methods. With these, we measure the perceived quality of motion cueing in specific trajectories (either recorded from real vehicles or generated in simulation) by comparing the actual physical motion of the simulator to features of the visually-presented motion inside the cabin. The picture below show the results of a validation experiment, in which four algorithms were compared.
Result of an experiment where two traditional MCAs (Classic and Classic+) were compared with two perception-based MCAs (Percept1 and Percept2). The "Score" axis shows the goodness rating of participants.