Wall-Following%20Research

1
Wall-Following Research

• Researcher Benjamin Domingo
• Mentor Joey Durham

2
Objective

• The objective of this research project was to
  build an autonomous robot that would be able to
  follow a wall.
• The robot would also have to maintain a certain
  distance away from the wall for which to follow.

3
Method

• Firstly, an intense analysis was performed in an
  effort to understand the robot perspective.
• An understanding of the robot environment was
  needed in order to proceed with developing a
  strategy for which to implement on the robot.
• An illustration of the environment for which the
  robot sees is shown in the following page along
  with a picture of the actual robot in the
  laboratory.

4
Figure 1. Robot environment
Figure 2. Autonomous robot
5
Environment
Sensor range

• A closer look at figure 1. tells us the point of
  view of the robot. It gives knowledge of what
  obstacles it may see and cannot see. The sensor
  data however, has some parameters. These
  parameters include having the ability of only
  reaching a 240 degree span, a max range of 5
  meters, there are only 683 sensor data point
  within the 240 degree span, and the frequency of
  the data is 10Hz.

6
Method

• Now that there was a clear understanding of the
  environment and the specifications of the robot,
  implementing a strategy was the next step.
• My idea for proceeding with the wall-following
  problem was to use the concept of damped
  oscillations and reaching a steady equilibrium.

7
Method

• An illustration of the damped oscillation concept
  taken from a text book is shown below .

8
Challenges

• Some of the challenges that were met were very
  difficult to handle.
• There were two reference frames that needed to be
  juggled at the same time. They were the local
  reference frame of the robot and the global
  reference frame of the environment.
• A thorough understanding of both reference frames
  and how they correspond to one another proved to
  be one of the most difficult challenges for which
  to overcome.

9
Challenges

• Another challenge included a comprehension of the
  computer language that was used to communicate
  with the robot.
• The use of C language and an understanding of the
  linux software Player/Stage was necessary in
  order for my ideas to translate onto the robot.
• Another challenge was that of left-handedness and
  right-handedness.
• For example, the robot had to choose whether to
  follow a wall that came within the 0-120 degree
  span or the 120-240 degree span. This problem
  would present itself when the robot would
  encounter two or more walls within its range of
  sensor data.

10
Results

• The result of much work in the detail of the
  programming and understanding the robot fully was
  successful to a certain degree.
• The robot incorporated the damping oscillation
  concept.
• The robot incorporated the idea of range gaps to
  deal with the problem of right-handedness and
  left-handedness.

11
Results

• Upon integration of my program in Player/Stage
  and running a simulation of it, the result is
  shown below. Here, the robot initially realized
  that it was too close to the wall. It needs to
  maintain a distance of 2 meters away and has
  corrected itself in the future development of its
  path.
• Note that since the wall in simulation is not
  smooth and contains ridges and bumps, the robot
  acknowledges these characteristics and imitates
  them exactly as a result.

12
Results

• Here we see that the robot has understood that
  there is a new wall to follow as it negotiated
  the turn on the upper right corner. This is a
  result of the right-handedness method.

13
Results

• In this screenshot, we can see that at this
  moment in time the robot is negotiating the
  incline of the wall it is following.

14
Conclusions

• From simulation it can be seen that this method
  of damped oscillations is a competent one.
• A next step in the development of this code is to
  have it evolve into a more robust form. Having
  the robot tackle more complicated environments
  where an erratic piece of data might show up is
  good next step.

15
Acknowledgements

• Professor Francesco Bullo
• Joey Durham, UCSB
• Edwards, Henry C. Differential Equations and
  Boundary Value Problems. Third edition. Pearson
  Education, Inc. 2004