Tactile graphics and raised-line drawings offer a rich domain for studying haptic perception, yet robust metrics of tactile dissimilarity between raised-line patterns remain underdeveloped. Pattern distinguishability is typically assessed individually, despite the local, sequential nature of tactile exploration and the importance of transitions between adjacent regions. To address this gap, we present a study comparing perceived transitions between adjacent tactile patterns with similarity judgments obtained from exploring the same patterns individually. The goal of this work is to probe the underlying perceptual organization that governs tactile pattern discrimination and continuity. Seventeen participants completed two complementary tasks under blindfolded conditions: (1) a free-sorting task in which patterns were grouped according to perceived similarity, and (2) a pairwise comparison task yielding confidence-rated binary judgments of transition saliency (i.e., how noticeable the transition between two adjacent patterns is) collected using a custom-built experimental interface. Results obtained from applying multidimensional scaling (MDS) to the free-sorting data converge with those of applying stochastic triplet embedding (STE) to pairwise judgments. Both approaches independently recover the same global structure. The alignment of the derived clusters across the two tasks provides initial empirical grounding for future work toward a relative metric of tactile dissimilarity, accounting for both individual pattern perception and local transitions.
Distinguishing rolling from sliding is necessary for realistic object manipulation in virtual environments. We study single-finger rendering of rolling vs. sliding using spatiotemporal electrical stimulation patterns and mechanical vibration. Motion cues are parameterized by contact display area, and the movement ratio between stimulus motion and finger motion, and DC-motor vibration is added as a texture and friction event cue. Two user studies show a clear perceptual disparity: narrow contact with a rolling-consistent ratio reliably elicits rolling, whereas sliding is ambiguous without event cues. High-frequency vibration robustly increases perceived roughness, reducing the unrealistic frictionless sensation of sliding rendered without vibration and improving physical plausibility. These results support a layered design in which geometry and kinematics establish motion mode and vibration supplies texture events for a robust sliding-related experience.
As digital interactions increasingly mimic physical reality, the integration of affective haptics becomes essential to mitigate sensory flattening. However, the perceptual hierarchy between vision and touch in emotional decoding remains an open question. In this work, we investigated visual capture and somatic fidelity in a trimodal emotion recognition context. Through a within-subjects experiment utilizing the RAVDESS dataset, we quantified response latency and error rates across congruent and conflicting audio-visual-haptic scenarios. Analysis of the results using linear mixed-effects models revealed a sensory hierarchy: visual cues consistently override haptic feedback. Crucially, our analysis uncovers a temporal asymmetry in error dynamics. Visual capture emerges as an automatic mechanism, in which participants aligned to the visual illusions rapidly and without hesitation. In contrast, correctly identifying haptic targets among visual conflict requires significant cognitive inhibition, manifesting as a measurable delay in processing time. These findings suggest that , within the tested paradigm, bodily-grounded haptic feedback is not sufficient to override visual misconceptions, and that haptic channels may require distinct temporal priming or intensity modulation to effectively compete with the visual system.
Mobile robots used in Mobile Encountered-type Haptic Devices (mETHDs) must remain near VR users during idle periods, that is, when the robot is not actively providing haptic feedback, to minimize haptic positioning delay. However, close proximity increases collision risk because immersed users are unaware of the robot’s presence. We address this trade-off between responsiveness and user safety through a Deep Reinforcement Learning (DRL) approach for adaptive robot positioning. Using collected motion-captured human trajectories, we train a policy that drives the robot to appropriate waiting positions during idle periods. In simulation, we compare the learned policy against static and heuristic approaches, and we further validate it on a real robot through a real-world deployment. Results show that learning-based positioning reduces positioning delay and improves safety, indicating a practical direction for responsive and safe mobile encountered-type haptics in large-scale VR.
