Snake Arm Glove Project Using Atmega32

For our project, we designed a glove that can be used to control the Cornell University robotic Snake arm thereby enabling a surgeon to remotely operate the snakearm as a colonoscope in conjunction with a vision guide system (aka TV goggles). The glove allows for movement of the arm’s segments as a whole and individually.
Snake Arm Glove Project Using Atmega32
More about the SnakeArm
A snake arm emulates the movement of a snake, with one point of attachment where the arm is anchored. It has multiple degrees of freedom, high flexibility, and thin diameter.
The current generation snakearm has been designed to be used in surgical applications � particularly colonoscopy. The need for a friendly control method emerged earlier in the course of our team efforts. The motion sensing glove is designed to meet this need by controlling the servo motors on the snakearm indirectly through the PC based on its movement.� The glove is composed of three types of sensors � flex sensors, a two axis accelerometer and several hall effect sensors. The flex sensors are used to select modes of operation, the accelometer is used to generally move the snakearm while the hall effect sensors are designed to enable precise control of the head
The vision guide system is the part that makes this project more appealing to surgeons. There is a camera mounted on the head of the snakearm itself and the camera is connected to TV goggles. So, the surgeon is able to control the snakearm with the glove while using the vision guide system to see inside the colon of the person being operation on. This is very much like playing a video game and this glove based system has been designed to enable remote colonoscopy surgeries.
For more information on the current snake arm capability, please see this youtube video.
II. High Level Design
Our idea came from our involvement with the snake arm team and the wish to control it in a more logical fashion. Having it mimic the movement of a person’s arm would allow them to have a more complete and powerful control over the arm allowing them to move it exactly as they please. The snake arm team is currently developing the snake arm for a surgical application, thus, accurate control is very important.
The project consists of the Mega32 serving as an interface between the hardware and the snake arm interface. Using an accelerometer we determined the user’s hand motion on the x and y plane, which controlled the base of the snake arm. The tilt determined the arm’s movement vertically, while the pitch/roll determined the movement of the arm horizontally. The flex sensors were used to activate and deactivate the hall effect sensors, which were used to control the head of the snake arm. The flex sensors and accelerometers went into the ADC of the Mega32, while the hall effect sensors go into PORTB. The inputs from the hardware components are calibrated and reformatted into two bytes that are sent across the serial port to a java interface that manipulates the snake arm.
One of the questions we asked ourselves earlier in the project was whether we should use send the data from the glove through the ATmega32 to the PC (hence snakearm GUI) or connect our ATmega32 directly to the snakearm servo motors and send out PWM signals to move them. However, keeping in mind that the snakearm is quite complex � just four segments of it amount to 12 motors and each motor has to moved in a perfect puppetering action with regard to the other motors to move a segment in a simple direction. Since a control system was already implemented for the snakearm in Java, it made no sense not use it and try to reinvent the wheel.
III. Program/Hardware Design
Hardware Details
The hardware we used consisted of the STK500 programming board, Mega32 microcontroller, a two axis accelerometer (Freescale MMA6261Q 1.5g accelerometer), 5 flexsensors (FLX01), and 8 Allegro 1101 Hall Effect Sensors. We also used a Freescale 9V battery to 3V converter, which served as the supply voltage to the accelerometer. The accelerometer’s x and y axis outputs tied into Pin A.0 and A.1. The flex sensors and hall effect sensors were both inputted into LM258/LM358 opamps and used a non-inverted style setup to amplify their voltage The hall effect sensors outputted digital signals that were amplified to 3.6 V, which is logic high. The sensors detect fluctuations in magnetic fields, so when a magnet is held next to one, the sensor goes low. The flex sensors also went into opamps to get a gain into a range of .5 – 2 volts, based on how much the sensor was bent. The greater the degree of bending the lower the output voltage. The output voltage is determined based on the equation Vin * R1 / (R1 + R2), where R1 is the other input resistor to the non-inverting terminal.
Program Details
We had one task that handled all of the inputs to the ADC and activity of the hall effect sensors. We first read the x-axis on the accelerometer from the ADC and then the y-axis(on the second iteration of the task) by incrementing the ADMUX variable. In the case that this was the first reading, these inputs were used to calibrate the readings from then on. We then got readings on the flex sensors in one iteration, by using a for loop that constantly incremented the ADMUX and polled the next port. We compared the reading on the flex sensor to a reference value. If it was less than that we determined that the flex sensor was being bent significantly. We used an array to store the pattern of flex sensors that were being flexed.
To actually get readings from the ADC, we used an ADC interrupt that gave us the latest conversion whenever it was ready. We used a sleep command when making the ADC read the next port, so that we could give it enough time to make the conversion to an accurate value. We noticed early on that we were getting random values from the ADC and after consulting the websites for previous mouse type projects we found that the sleep command could be used to help give the ADC ample time to make the conversion.
The id of the hall effect sensor that was activated was only transmitted when all of the flex sensors were being bent. Another interrupt was used to determine which of the flex sensors were being activated.
Sending the data across the serial port was done in three bytes. The first byte was merely a buffer byte, so that the Java interface could determine the order in which the bytes were being sent. The second byte consisted of the flex sensor array and whether or not the hall effect sensors had be activated. The third byte consisted of the reading from the accelerometer or the identity of the hall effect sensor that was being used. The upper nibble of the byte stored a 3 bit value that ranged from 0 to 4, which determined how quickly the user wanted to turn on the x-axis. The msb bit was used to determine direction. The lower nibble contained identical information on the y-axis. The byte would contain the ID of the hall effect sensor that was activated if they were being used. Our java interface opened a line of communication with the microcontroller and polled the serial port for data. We used queues to handle the quick stream of data and extracted the necessary data from them, before sending them to the snake arm interface. The snake arm is controlled by a program that communicates to the servo motors via the ATiny26. The program gets feedback from the ATiny26, so we obtained the current position of the snake arm and updated to where we would like it to move based on our readings from the accelerometer and hall effect sensors. This last part of interfacing was done with the assistance of one of the CS members of the snake arm team. The code for the microcontroller and serial interface with microcontroller is provided. However the snake arm interface could not be provided and is not really necessary for the user to see anyway.
1st byte – 1100000000 (buffer)
2nd byte – 2 bits – not used, 5 bits (Flex sensor array) 1 bit (Hall Effect or Accelerometer)
3rd byte – 4 bits x axis acceleration 4 bits y axis acceleration or 8 bit id of hall effect activated

These three bytes are sent through the UART (serial) interface to the computer where a Java program picks them up ( and interprets the data. A program runs as a new thread so that it takes up less CPU power while polling the COM port as necessary. Next, this data from this program (sensor positions) are extracted by the CU SnakeArm GUI which runs them through its programmed control system and moves the servo motors controlling the snakearm accordingly.
Snake Arm Glove Project Using Atmega32 schemetic
Looking back at the program implementation sector of our project, the serial communication and interfacing with the snakearm GUI and control system proved to be quite tricky. There were numerous errors that we encountered from bad values, inconsistent formats, etc. Moreover, integration this into an existing yet functioning platform proved to be demanding.
IV. Results of the Design
We adhered to the standard format of communication over the serial port. We used a baud rate of 9600 bps, sent 8 data bit packets, had 1 stop bit, and no flow control or parity bits. Although we were receiving at this rate on the computer end, we only handled data every half a second, dropping all other packets. The reason for this was to avoid overdriving the servo motor. The response went down as a result, but is still acceptable. This was needed for the safety of the snake arm. We also enforced safety in the glove by keeping the flex sensors within one of the layers of the glove and keeping the accelerometer on small breadboard on top of the glove to give it a proper shielding from any static that could form on the glove. Our project does not interfere with any other devices in any way. Overall, the snakearm does respond to our glove and the sensor readings on the GUI correspond properly to our sensors. However, the movement is not very smooth. This is a restriction from the existing control system which we did not create or modify in this project.

 Parts List:




Part Number

Stk500 + Mega32

Free (SnakeArm team provided)

Not needed

Flex Sensor (5)

$12.95 * 5 = $64.75

Imagesco: FLX-01


Free (Sampled from FreeScale)

FreeScale: MMA7260QT


11*$1.00 (Lab Fee) = $11.00




Not Needed

Hall Effect Sensors (8)

Free (sampled from Allegro)

Allegro A1101EUA-T

Jumper Cables

2*$1.00 = 2.00

Not Needed

Several Breadboards

Free (SnakeArm provided)

Not Needed

Freescale 9V to 3V battery converter

Free(sampled from freescale)

Not Needed



For more detail: Snake Arm Glove Project Using Atmega32

About The Author

Ibrar Ayyub

I am an experienced technical writer holding a Master's degree in computer science from BZU Multan, Pakistan University. With a background spanning various industries, particularly in home automation and engineering, I have honed my skills in crafting clear and concise content. Proficient in leveraging infographics and diagrams, I strive to simplify complex concepts for readers. My strength lies in thorough research and presenting information in a structured and logical format.

Follow Us:

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top