Many VR applications, particularly those involving training or teleoperation, require realistic object manipulation. Current VR systems often limit object orientation to the user's hand movement, lacking the nuanced control afforded by varying grip strength in real-world interactions. This paper focuses on replicating the ability to adjust grip strength and its influence on object rotation using force-feedback gloves. The human hand's dexterity allows for both extrinsic (hand movement) and intrinsic (finger manipulation within the hand) control. The force applied by fingers significantly impacts how an object is held, allowing for actions like letting an object slip or controlling its rotation subtly. While previous research explored pressure-based input for specific actions (e.g., squeezing a smartphone, pinch gestures), this work builds upon earlier research using controller-based systems to control grip strength and the resulting object rotation. The authors' prior work demonstrated advantages in user satisfaction and realism using controllers, but this system introduced the need to map button presses to virtual actions, increasing mental demand. Subsequent research with Valve Index controllers improved input fidelity by directly interpreting pressure on the controller handle as grip strength. This study aims to explore the potential of leveraging the precise hand pose and finger input capabilities of force-feedback gloves in combination with haptic feedback to achieve even more natural and intuitive object manipulation.

The paper reviews existing research on haptic devices providing kinesthetic forces for hand and finger interactions, focusing on translational manipulations, shape exploration, and weight simulation. A survey of glove-shaped haptic devices highlights various designs and their characteristics. Past work using squeezing actions on haptic devices has been explored in abstract ways (e.g., smartphone interaction, pinch gestures in mixed reality) and for basic object manipulation. Previous research also explored object elasticity in VR through pressure with controllers or force-feedback gloves. The authors expand on their prior work using controller-based systems to manipulate virtual objects with variable grip strength, allowing for both fixed and freely rotating object states based on grip pressure. This approach avoids the need for repetitive clutching (releasing and regrasping), enhancing interaction fidelity. The advantage of variable grip in various everyday tasks is highlighted, showing that the potential for more realistic interactions in VR could benefit use cases demanding high fidelity.

A user study (N=21) was conducted using the SenseGlove DK1 force-feedback gloves and an HTC Vive Tracker for hand position tracking. The system tracks finger pose and movement, using cable-based brakes to simulate contact force. The maximum force applied at each fingertip is 40N, updated at 200Hz. The virtual environment was created using Unity, simulating a workshop setting with photorealistic objects. The interaction system maps the glove's resistance force to pressure against the virtual object. A copy of the object is created upon grasping, allowing the copy to rotate freely (if the grip is loose) while maintaining force feedback from the original object. The object's rotation is anchored to a point between the thumb and index finger. Initially, a continuous grip strength control based on all fingers' pressure was attempted, but this proved unreliable. The system was simplified to binary grip states (firm or loose) based on thumb and index finger pressure, visualized by a color-coded bar above the hand. Two tasks were used: Task A (repeated movements of six cans with identical start and target poses) and Task B (movements of twelve objects of three different types—cans, books, and milk cartons—from various start poses to a single target pose). The study was within-subjects, comparing fixed grip (object rotation locked to hand rotation) and variable grip conditions. The NASA-TLX questionnaire, Presence Questionnaire, and custom questions assessed task load, presence, and perceived control. Data were analyzed using the Wilcoxon signed-rank test due to non-normal distribution. Semi-structured interviews provided additional qualitative data.

The results showed that participants performed significantly better in the fixed grip condition across several metrics. Translational accuracy (distance to target center) was higher with fixed grip (median 8 mm vs. 9 mm for variable grip), though the difference was small. Rotational accuracy (deviation from target orientation) was also better with fixed grip (median 4.2° vs. 6.1°), again with a small effect size. However, participants took significantly longer (median 3.8 s vs. 5.2 s) and required more grasping attempts (median 1 vs. 2) to place objects successfully in the variable grip condition, with large effect sizes observed. The fixed grip condition was rated better in several questionnaires. While the overall NASA-TLX score did not show a significant difference after correction, the subscales of mental demand, performance, and frustration indicated higher demands for the variable grip condition. The Presence Questionnaire indicated significantly better interface quality ratings for the fixed grip condition. Custom questions showed significantly better ratings for moving and rotating items as expected in the fixed grip condition. The interviews revealed user struggles handling the glove itself, along with unexpected or absent haptic feedback.

The study's unexpected results—poorer performance with variable grip compared to expectations and previous controller-based studies—can be attributed to several factors. The SenseGlove DK1, as an early prototype, presented challenges: its force feedback mechanism is sensitive to finger angle, leading to unpredictable resistance. The limited ability to accurately measure finger pressure also contributed to difficulties in controlling grip strength. Many participants were novice VR users, further impacting performance. In addition, accidental dropping of objects was more frequent in the variable grip condition. The visual grip strength indicator was perceived as distracting. Finally, the lack of cutaneous cues (frictional sensation at the fingertips), which are crucial for precise grip control in real-world interactions, may have contributed to the difficulties. The findings do not completely dismiss the concept of variable grip using force-feedback gloves, but they highlight the need for improved hardware and haptic fidelity.

The study presents a system for controlling virtual object rotation via finger pressure using the SenseGlove DK1, but the evaluation revealed poorer performance and increased task load compared to a fixed grip. Technical limitations and the absence of cutaneous feedback highlight the need for improved haptic interfaces. Future work should focus on incorporating tactile cues (shear forces, friction, etc.), more accurate pressure sensing, continuous grip control, and improved physics simulation to create more intuitive and realistic object manipulation in VR. The variable grip approach remains promising, especially in applications where gravity is not a factor.

The study's sample included many novice VR users, which may have influenced the results. The SenseGlove DK1, being a prototype, has limitations in force feedback and pressure sensing accuracy. The visual grip strength indicator proved distracting. The lack of cutaneous haptic feedback is a significant limitation, impacting users' ability to achieve a precise grip. The binary grip implementation, while necessary due to system limitations, may not fully capture the nuance of continuous grip strength adjustment.