A Room to Roam: Reset Prediction Based on Physical Object Placement for Redirected Walking

TL;DR

This work tackles predicting and reducing resets in Redirected Walking by leveraging a Vision Transformer to estimate resets from top-down layouts of physical spaces, trained on simulation data. It introduces an interactive Unity-based interface that provides real-time feedback as users rearrange furniture, enabling rapid interior planning to minimize resets. Study 1 demonstrates that object placement significantly affects resets, while Study 2 achieves RMSE $= 23.88$, MAE $= 15.36$, and $R^2=0.91$, validating the predictive capability and its practical utility. Overall, the approach offers a fast alternative to time-consuming simulations or multiple user studies, facilitating more immersive RDW experiences within real-world spatial constraints.

Abstract

In Redirected Walking (RDW), resets are an overt method that explicitly interrupts users, and they should be avoided to provide a quality user experience. The number of resets depends on the configuration of the physical environment; thus, inappropriate object placement can lead to frequent resets, causing motion sickness and degrading presence. However, estimating the number of resets based on the physical layout is challenging. It is difficult to measure reset frequency with real users repeatedly testing different layouts, and virtual simulations offer limited real-time verification. As a result, while rearranging objects can reduce resets, users have not been able to fully take advantage of this opportunity, highlighting the need for rapid assessment of object placement. To address this, in Study 1, we collected simulation data and analyzed the average number of resets for various object placements. In study 2, we developed a system that allows users to evaluate reset frequency using a real-time placement interface powered by the first learning-based reset prediction model. Our model predicts resets from top-down views of the physical space, leveraging a Vision Transformer architecture. The model achieved a root mean square error (RMSE) of $23.88$. We visualized the model's attention scores using heatmaps to analyze the regions of focus during prediction. Through the interface, users can reorganize furniture while instantly observing the change in the predicted number of resets, thus improving their interior for a better RDW experience with fewer resets.

Figures (6)

• Figure 1: Overview of our reset prediction model. Our model predicts the number of resets an RDW user will experience in the given physical space based on the top-down view. The information of objects placed in the physical space is first transformed into a binary top-down image, filled with 1s where objects are present and 0s where they are not. The image is fed into the Vision Transformer (ViT) model, where it is divided into patches of size 16 $\times$ 16 and embedded into a 768-dimensional space using linear projection. The image passes through 12 encoder layers of the transformer, and the number of resets is predicted through the Multi-Layer Perceptron (MLP) head. The data used to train the model was collected through simulation.
• Figure 2: The figure shows one of the simulation environments in which the data collection was carried out.
• Figure 3: An illustration where the number of resets changes depending on the placement of obstacles in the physical space. The top-down view displays objects in white and empty space in black.
• Figure 4: Mean and standard deviation ($SD$) results of the number of resets as a result of placing the same number of obstacles in physical space. Error bars indicate 95% confidence intervals.
• Figure 5: Scatter plot illustrating the performance of the trained model on both the validation and test sets. The blue points represent the combination of the actual number of resets ($x$-axis) and the predicted number of resets ($y$-axis) for each data point. The red line indicates where the predicted values would be exactly the same as the actual values.