Problem Report

Home
click here to download Word file

Artificial Intelligence and Robotics

iRobot Create

Jose D. Garcia Gopal Jani Amit Ramani

I.Introduction
A. Objective
B. Equipment and Software Installation
C. Downloading your code to the iCreate Robot
II. iCreate Robot Algorithm
III. iCreate Robot Demos
A. Demo One:
1. Problem Faced:
2. Resolution:
B. Demo Two:
1. Problem Faced::
2. Resolution:
B. Demo Three:
1. Problem Faced::
2. Resolution:
B. Demo Four:
1. Problem Faced::
2. Resolution:
IV. Important Factor for Future Work

Introduction

Objective

The purpose of this project is to program an iRobot Create series robot to map the Computer Science Department corridor and perform the following tasks:

First, the robot traverses from the Computer Science second floor’s elevator and knocks on Rosely’s office door to pick up some cookies.
Second, the robot then attempts to knock on Dr. Silaghi’s door to let him know the cookies has been picked up.
Third, the robot then attempts to deliver the cookies to the final destination which is Dr. Stansifer’s office door.
Finally, the robot needed to do the first three tasks in the shortest amount of time possible to win the race against the other team’s robots.

Figure 1-Map of Computer Science Building second floor

Figure 2-Map of path followed by robot to mapped the floor

The robot make use of its sensors to locate itself inside the corridor and make decisions such as when to make a turn or as when just to keep going straight. Our algorithm revolved around sensing the walls with the robot’s bump sensors and keeping track of the distance of the walls to minimize uncertainty to whether we are at the desired location or not. However, there are pros and cons for using this kind of method which will be explained later.

Equipment and Software Installation

Our team used the final program as our baseline from the previous work of students who took the class in the spring of 2009. They used the python programming language to implement their project. The reason why they picked python was because python have an easy and user-friendly serial communication interface to communicate with the robot. Also, a Create package programmed in python, available for download from Dr. Zachary Dodds’s personal website, provides all the interfaces needed to control the robot’s actuator and as well as access to its entire sensor’s data.

To set up your system to run our source code on the iCreate robot, you need to do the following:

Install Python 2.4 - (Download Here)
Install PySerial for Windows – (Download Here)
Install Win32All for Python 2.4 – (Download here)
Install Create Package (Download Here)

Once you have installed the files above, you are ready to download our code (or your code) to the iCreate robot. PySerial assume that you have serial port (RS-232) in your laptop for communication. However if you do not have a serial port but have a USB port, you are going to need a USB-to-Serial cable. We bought a USB-to-Serial cable from RadioShack for $39.99. You can buy a cheaper one online for about less than half of that price. Here is a link to the cable our team used for the project (USB-to-Serial Cable).

Figure 3-Gigaware® 6-ft. USB-A to Serial Cable (RadioShack)

Downloading your code to the iCreate Robot

To download you code to the iCreate robot, follow the following steps:

Connect the USB end of the USB-to-Serial cable to your laptop (if the laptop you own do not have a Serial port) and the other end to the serial port of the iCreate Robot cable.
Verify in the Python code, the right port is specified for communication.
To run the program, Type the following commands in the Python command line:

Type “ import Create ”
Type “ from Create import* ”
Type “ import nameofpackage ”
Type “ from nameofpackage import* ”
Type “ aimain() ”

ICreate Robot Algorithm

In this section, we are going to describe the main algorithm to program the robot:

ICreate Robot Demos

Demo One:

The task assign in first demo was to Start from Computer Science second floor’s elevator and knocks on Roselyn’s office door to pick up some cookies.

Problem Faced:

At the time of first demo, the planning for mapping the corridor of computer science building, program was based on the traveling time from one point to another, the distance covered by the robot and the angle at which it will turn. But by considering these three factors it was not possible to have a robust program through which the robot can reach to the destination. We have faced so many problems in that program and that problems are as follows,

Time factor depends on the robot’s battery charging level. That means, consider the case where the charging of the battery is full, in that case the robot will go faster because of the rotation of the motor for the wheels will be faster, and I case when the battery level is low the rotation of the motor will be slower than normal. So for both the cases we will go forward for particular seconds but the distance covered by the robot will be different.

Distance is also about the same factor as the time because as we have explained in time, distance is also not reliable factor. On each run the distance covered by the robot for same path will be different.

Angle of turning the robot in particular direction will depend on the previous position of the robot and main importantly the previous angle may differ because of the surface of the floor. So we cannot determine how much was the previous angle and at how much angle we should make turn.

Resolution:

To solve the above mentioned problem, we decided to redefine the algorithm and we decided to develop a new strategy. The idea in the new algorithm was to follow the wall by left side wall sensor of the I-create robot. In this strategy we were not bothering about the time and distance. We programmed the robot that will follow the left hand side wall and by identifying the points it will map the path and in this way robot will identify where it is while reaching to the destination.

Demo Two:

Phase 1:
The task assign in Phase 1 of second demo was to Start from Computer Science second floor’s elevator and knocks on Roselyn’s office door to pick up some cookies and reach middle of the hall way without obstacle.

Phase 2:
The task assign in second demo was to Start from Computer Science second floor’s elevator and knocks on Roselyn’s office door to pick up some cookies and reach middle of the hall way with obstacle.

Problem Faced:

As mentioned, robot was following the left hand side wall to search the path and go through that. We have faced problems in that program

The angle at which it will turn little bit in right side when it will touch the wall was not working as intended it was going in wrong direction. It was working good when it will take right turn but when it was trying to take the left turn again to sense the wall it was not able to compensate the angle and due to that it was going so much right that it was hitting the wall with the front bump and the algorithm was failing.

Resolution:

To solve this particular problem we reduced the speed while turning. Due to the reduced speed the robot was able to take sharp turn in specified directions. And the problem was fixed up to a level. Though reduction in speed solved the problem but still it was not a reliable solution it needed a better solution which we figured it out later.

Demo Three:

The task for the Third demo was to Start from Computer Science second floor’s elevator and knocks on Roselyn’s office door to pick up some cookies and reach Dr. Silaghi’s office door.

Problem Faced:

At the time of the third demo there was lots of problem because there was a new task as well as there was a new milestone, that mile stone was obstacle in the path on which robot will travel. There was issue regarding the angle of turning which we have discussed. Due to this we have again refined the algorithm and coded it accordingly to make it robust. All together the problems faced was as below,

The new milestone of having obstacle in the path was difficult to handle, because for that we made so many changes in the algorithm, program flow as well as in the code.
Secondly as we were already aware of the problem of taking turn right and left while following the wall. So at that time it was very important to solve this problem.
There was a bug in the code.

Resolution:

To solve the condition in which the robot will have the obstacle anywhere in the path we took distance in the picture. We added logic in the following wall and that logic was to keep track of distance. We were aware of the thing that the distance factor is not reliable. But by keeping track of minimum distance from one point to another and adjusting that distance by adding some offset we managed to have more reliable reading of distance. After that the task was to applying that logic in the running condition. For that we have given the condition that specifies that if the distance is less than the minimum and robot hits the front bump, there is an obstacle, and if the distance is greater than the minimum distance then it is the point where the robot was intended to reach.
Now the challenging task was to solve the problem of turning in right and left while following the wall. Just reducing the speed it will not solve the problem because we was making robot to turn in specified angle. To solve this we added code that will not specify the angle. After hitting the wall left wall sensor will be activated so we specified minimum hit to the wall by left bump and if hit to the left wall by bump is greater than the minimum, it will stop, and it will turn in to right direction until it will loses the contact to the wall and the left bump. And it will go forward with some degree angle to the left side to find the wall again.
After doing lots of run we came to know that the bug due to one unnecessary call to a function and after removing that function call the problem was solved.

Demo Four:

In final Demo the task was to Start from Computer Science second floor’s elevator and knocks on Roselyn’s office door to pick up some cookies and reach Dr. Silaghi’s office door and then final reached Dr. stansifers office door.

Problem Faced:

At the time of final demo the task was more difficult because there were 3 obstacles in the path. But actually that was not the problem in our case because at first time when the obstacle was introduced, we developed the code that can overcome as many as obstacle in the path and anywhere in the path. So the problems faced this time where as follows

We got one problem which we never got before, and that was, when it was hitting the front bump to the wall while detecting the wall it was intended to stop and turn parallel to the wall in right direction. But what was happening, it was stopping after hitting the wall and after that instead of taking right turn to be parallel to the wall, it was taking turn in some weird direction. In this case it was also difficult to find out what exactly is going wrong because this kind of thing was never happened. We found this same problem at three different places.
Another problem was a normal problem with the calculation of covered distance. It was a simple error so we fixed it immediately as we found it.
Moreover, we found that at one place it was detecting as reached to the door but it was not at the door. That was also a normal problem which was very easy to figure it out and solve it.

Resolution:

After so much work we actually figure it out that why the problem was occurring though it was not occurred regularly. So the problem was, after hitting the wall the robot was not able to reset the sensors and it was not compensating from the effect. Due to that the angle of turning was getting massed up. To solve this we have simply applied some logic where after hitting the wall it will stop for some time and go little bit reveres to compensate the sensors and the angles. In this way the problem was solved and worked very well without any error or any uncertain results.
We have just recalculated the approximate distance and putted it down in the code and the problem was solved immediately.
This problem is almost same as previous problem so its solution was also same. So we fixed it up very quickly.

Important Factor for Future Work

The ICreate robot can be programmed in very efficient manner using the sensors it has. But for the future work we have to keep in mid following factors.

Behavior of the robot depends on the battery level of the robot (but it is not a big problem in this case behavior will be changed negligible).
It is also dependent on the computer system you use. Because in each and every computers the computational time is different. So our suggestion is to use same laptop if possible because it affects a lot in robot behavior.
Sensors in ICreate robot do not give the value as just 0 and 1 or say true or false. It provides a range of signals like 0, 1, 2, 3, 4, 5…
It is also possible that it will give different reading of sensor at same place, because it depends on the battery level.