🤖Soft Neck Robot

This repository contains my Master Thesis project, in which I had to control the bending of a Soft Neck Articulation through the feedback of a Soft TPU/CB Flexible Sensor.

The research has been held at the Robotics Lab at Universidad Carlos III de Madrid: https://softroboticslab.gitlab.io/website/index.html

The Soft Sensor requires the multimeter or the Arduino in order to be sensed, according to the chosen code.

Multimeter

Multimeter Setup

1. To connect the multimeter, check the presence of the code multimeter_communicationLinux.cpp in the Soft_neck_Soft_sensor/main folder: this code is compatible with every OS since it is based on the SerialComm.h library.
2. Press Power On
3. Switch to "Four-Wires Mode": a) Press Shift and then b) Press 4W

Multimeter Common Issues

1. ERROR displayed during multimeter initialization

   The code automatically set the multimeter to LOCAL mode and then to REMOTE mode. When the code execution crashes or is forced to quit, this action may generate an error during the following test. So after an interrupted test, shut down and restart the multimeter before starting a new one.
   

Arduino

Arduino Setup

1. Open the Arduino IDE
2. Modify Filter.ino in Soft_Neck_Soft_Sensor/Filter folder with the correct pinout, baud rate, filter parameters, resistance value (of the Tension Divider Circuit).
3. Upload Filter.ino on the Arduino

Arduino Common Issues

1. Arduino not detected

   • Arduino IDE should be opened before running any code on Qt Creator: this helps initializing the connection between PC and Arduino.
   • Check that the Serial Monitor and Serial Plotter of the Arduino IDE are closed: otherwise the data is sent to them instead of Qt Creator.

2. Wrong resistance output

   Check that the fixed resistance (of the Tension Divider Circuit) in Filter.ino is the same used in the breadbord and check that the cable connections of the breadbord are correct.

General

Setup and Recommendations

1. Neck's tendons need to be manually tightened before any demo without the neck bending in any direction and avoiding to compress the soft link (for a perfect vertical position help yourself with a bubble level). Do it with the motor being shut down and be sure to tighten clockwise direction: the tendon should be on the outside.

   

2. Check that every cable is connected: power, CAN bus, IMU and Arduino/multimeter.

3. Check in the chosen code, for IMU and Arduino/Multimeter, that the specified Serial Ports are the same of the actual used ones. For IMU and multimeter you check it from the Terminal with the command:

   dmes | grep ttyUSB

   For Arduino you check it with the command:

   dmes | grep ttyACM

4. When running the code, be ready to shut down the motors in case of any control issue during the tests to avoid breaking the soft link.

Running the Code

1. Build with CMakeLists.txt in the outer Soft_Neck_Soft_Sensor folder (it will call all the CMakeLists.txt of the sub-folders automatically building every source file).

2. Choose your code in Soft_Neck_Soft_Sensor/main folder and open it through Qt Creator:

   Name	Multimeter/Arduino	Goal	Feedback
   9.cpp	Multimeter	Control	IMU
   10.cpp	Multimeter	Sys Identification	Open Loop
   14.cpp	Arduino	Control	FlexSensor
   15.cpp	Arduino	Sys Identification	Open Loop

Generating New Reference Signals

In each code, after IMU calibration commands, there is a section dedicated to the generation of Reference Signal introduced and followed by a line of ****.

In this section, reference vectors can be created by:

1. generating simpler child vectors as "target0", "target1" with ad hoc methods, being able to specify: amplitude, rate, duration, number of cycles, frequency...
2. concatenating child vectors to a parent vector like "target_pitch" (target_pitch = target0, target1, target2, target3...)

std::vector<double> target_pitch;

std::vector<double> target0;
referenceCyclicRamp(target0, dts, 0.35, 0.17, 3);
target_pitch.insert(target_pitch.end(), target0.begin(), target0.end());

std::vector<double> target1;
referenceStep(target1, dts, 0.52, 20.0);
target_pitch.insert(target_pitch.end(), target1.begin(), target1.end());

ATTENTION: code 14.cpp also contain a preload_pitch refence signal that is tracked in position OPEN-LOOP before target_pitch (preload_pitch = preload0, preload1, preload2...). DO NOT COMMENT it unless you are sure that your flex sensor does not present any initial drift. If the drift persist for more time, you can increase the duration of the preload motion.

Saved Data

• The data are automatically saved as .csv files in the Soft_Neck_Soft_Sensor/CSVdata folder after each test. The folder is not emptied before the saving, so be careful to not use data saved from previous tests.
• If you want to save more variables, check if you want to save a vector or a 2D vector: choose between the methods exportVectorToCSV() and export2DVectorToCSV().

Common Issues

1. Executions stops before IMU calibration

   The IMU could not close correctly every time, so the execution of the program may stop, usually printing this last line:

   Port /dev/ttyUSB0 has been correctly initialized

   To solve it, just Force Quit and RUN again.

2. Loose tendons

   After many tests or with a new neck, tendons can get loose and tangle (unknown reason, motor wear? asymmetric system?), leading to position drift and saturation. To solve it you can go to SoftCC3Tendon.cpp file in Soft_Neck_Soft_Sensor/lib/robot-device/robots containing the robot kinematics, and change the last lines of the SetRollPitchArcLen() method:

   if(tendons[0]<0) tendons[0] = 1.1*tendons[0];
   if(tendons[1]<0) tendons[1] = 1.1*tendons[1];
   if(tendons[2]<0) tendons[2] = 1.1*tendons[2];

   Since the tendons get loose only during negative velocities, we want to increase their absolute value only when negative (1.1 = increased by 10%, 1.3 = increased by 30% ...). Play with this amplification value until you have no more drift.

   For new necks, set the correct soft link height arc_length.