Virginia Tech
A Soft Capacitive Tactile Sensor for Force Feedback in Vision-Based Robotic Manipulation
Abstract
dc:description.abstractRobotic pick and place of fragile objects, such as ripe produce, is difficult not because of motion accuracy but rather force resolution at the moment of contact. The force estimates from a joint force sensor on a commercial robot arm cannot reliably halt a descent before the contact force exceeds the extremely low force thresholds of items like ripe tomatoes and strawberry slices. There exist soft tactile sensors that can resolve contact forces in the millinewton range, but they are packaged in a form that is difficult to integrate with other soft robotic end-effectors, such as octopus-inspired adhesives, which provide compliant contact at the interface for safe handling but lack the required force feedback to perform manipulation. This thesis presents a soft capacitive tactile sensor and an octopus-inspired adhesive end- effector module that close this gap. The sensor uses an Ecoflex 00-30 dielectric, a stretchable SIS/Ag-flake/EGaIn LM composite electrode pair, and an embedded plasma-treated PTFE electret that boosts low force sensitivity. A bare sensor calibration was performed and yielded a capacitance to force response that is well fit by a power law, which was read with a custom AD7745-based readout board. This board allows a resolution as low as 20 mN with a worst case SNR above 27 dB across all the trials performed for validation. The combined sensor and octopus-inspired adhesive module is mounted on a Franka Emika Research 3 robot arm and used to pick and place objects such as a rigid block, a ripe tomato, and a strawberry slice. To perform the manipulation task, a vision pipeline consisting of a Gemini based open vocabulary model paired with a state machine controller is used. Across the three object classes, which ranged from rigid to extremely fragile, the proposed sensing modality reduced the peak descent force overshoot into the range of millinewtons and increased the success rate from 0% to 100% on the two fragile objects (tomato and strawberry slice), showcasing that low-force tactile feedback is the missing element for vision-based robotic manipulation of fragile objects.
Degree
thesis:*- Name thesis:degree_name
- Master of Science
- Level thesis:degree_level
- masters
- Discipline thesis:degree_discipline
- Mechanical Engineering
- Department dc:contributor.department
- Mechanical Engineering
- Grantor dc:publisher
- Virginia Tech
- Year dc:date.issued
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Balusu, Sai
- Chair dc:contributor.committeechair
-
- Bartlett, Michael David
- Committee members dc:contributor.committeemember
-
- Li, Suyi
- Losey, Dylan Patrick
Subjects
dc:subject × 6Rights
dc:rights- Statement dc:rights
-
- Creative Commons Attribution 4.0 International
- Licence dc:rights.uri
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Dc Identifier Other
- vt_gsexam:47335