Force-Sensing Book - Manipulandum
Bachelor--Thesis at group TAMS
Motivation
Human manipulation skills rely to a large degree on force
and tactile sensing from the fingertips.
Measuring and modeling those forces during object manipulation tasks
remains an important open research question.
Unfortunately, those forces are very difficult to measure,
as external sensors on the fingers (e.g. a dataglove) restrain
the fingers and soften the tactile perception.
An alternative approach is to put the sensors into the manipulated objects,
so that the human motions are not disturbed and can be recorded.
The goal of this Bachelor thesis is to design and build an
instrumented multisensor object in the shape of a typical book,
that can be used to record and analyze a variety of common
manipulation tasks:
- pushing a book across a table,
- extracting a book between other books from a shelf,
- picking a book from a table,
- opening the book and turning a page,
- regrasping with both hands,
- etc.
Hardware design
To track object motion and to measure the applied finger forces,
the instrumented object will be constructed from several parts
connected via force sensors (either strain-gauges or optical sensors)
and joints.
At least one IMU will be included to track the overall object motion.
An integrated microcontroller (e.g. Arduino nano IOT or ESP 32)
will sample the sensors and transmit realtime data via WiFi
or save the data to a sdcard for offline analysis.
As the system needs to be battery-powered, a suitable
power-supply circuit is also part of the hardware design.
Details of the hardware design are up to discussion.
A modular construction with small and compact electronics
and exchangable passive outer shells to model books of different
sizes and weights would be best:
- WiFi/bluetooth microcontroller (e.g. Arduino Nano IOT, ESP 32),
- battery power (e.g. 3.7V LiPo with micro-USB charger,
- IMU for object pose tracking (accel, gyro, mag, 100Hz+),
- multi-part outer shell with embedded force sensors to
measure applied finger forces,
either strain-gauges (with HX711) or optical proximity sensors,
- optionally, tactile skin to track finger positions,
- optionally, vibration/deflection pads to track sliding motions.
Software and ROS integration
The microcontroller firmware will be as simple as possible,
sampling the on-board sensors and transmitting data to
the laboratory workstations for analysis and visulation.
To integrate with our ROS-based robotics and motion-tracking setup,
a small ROS node (python or C++) is needed to parse the
incoming WiFi data and publish the data to the ROS environment.
A 3D graphical visualization of object pose
and applied forces (e.g. arrows) could also be part of the thesis.
Thesis Goals:
- designing and building a multi-sensor embedded system
- hardware component selection to be discussed
- microcontroller firmware for sensor readout and WiFi transmission
- basic host software (python,C++) for data analysis and visualization
- collect ideas and experiment data leading to a publication
Requirements
- as always, interest in the topic area
- interest in building your own hardware
basic knowledge of electronics and hardware (e.g. Embedded Systems course)
- interest in human manipulation
Contact