Tactile-Based Human Intent Recognition for Robot Assistive Navigation
Shaoting Peng, Dakarai Crowder, Wenzhen Yuan, Katherine Driggs-Campbell
AI summary
Problem
Existing robot assistive navigation interfaces rely on joysticks, voice, or biosensors that are often unintuitive, distracting, or require extensive calibration, failing to replicate the natural physical guidance of human caregivers.
Approach
The authors mount a cylindrical tactile skin on a robot handle to capture grasp patterns and introduce the Cylindrical Kernel SVM (CK-SVM), which accounts for the sensor’s 3D geometry to robustly classify navigation intents despite natural rotational shifts in user grip.
Key results
- CK-SVM achieves 97.1% simulated and 90.8% real-world classification accuracy, outperforming four baselines
- Integration of a knitted cylindrical tactile sensor with a mobile manipulator for real-time intent recognition
- Pilot study confirms strong user preference for tactile control over joystick and voice interfaces
- Geometry-aware kernel explicitly resolves rotational grip shifts that degrade standard classifiers
Why it matters
Offers a natural, robust, and user-preferred tactile interaction paradigm that can significantly improve the safety and independence of assistive robots for mobility-impaired populations.
Abstract
Robot assistive navigation (RAN) is critical for en- hancing the mobility and independence of the growing popula- tion of mobility-impaired individuals. However, existing systems often rely on interfaces that fail to replicate the intuitive and ef- ficient physical communication observed between a person and a human caregiver, limiting their effectiveness. In this paper, we introduce Tac-Nav, a RAN system that leverages a cylindrical tactile skin mounted on a Stretch 3 mobile manipulator to provide a more natural and efficient interface for human navi- gational intent recognition. To robustly classify the tactile data, we developed the Cylindrical Kernel Support Vector Machine (CK-SVM), an algorithm that explicitly models the sensor’s cylindrical geometry and is consequently robust to the natural rotational shifts present in a user’s grasp. Comprehensive experiments were conducted to demonstrate the effectiveness of our classification algorithm and the overall system. Results show that CK-SVM achieved superior classification accuracy on both simulated (97.1%) and real-world (90.8%) datasets compared to four baseline models. Furthermore, a pilot study confirmed that users more preferred the Tac-Nav tactile interface over conventional joystick and voice-based controls. Code and video are available at: https://sites.google.com/view/tac-nav/home.