"Involving students from different disciplines in the discovery process, and exposing them to the joy of hands-on experience"

 

TeamBots

Home Mission Program Projects Robotics People K-12 Outreach Links
 
 
 
 

Robotics Academy Summer Research Project
Sensor Based Motion and Wireless Robot Communication

By Addie Sutphen

Initial Goals (5/19/03)
The project goal was to build a team of robots that work together, sharing sensor information to reach a common goal or target. The project goal was for the robots to constantly “share” their sensor data with each other. Each robot would then, using information obtained from the other robots, independently decide what sort of behavior action it should take. The initial goal was to create a team of five working, communicating robots by the end of the summer project.
The underlying concept of the robotic team comes from the development of an “artificial nose,” a fiber optic sensor that identifies any vapor or “smell” by its chemical signature. This artificial nose is not yet portable enough to ride piggyback on a Lego robot, so a Lego sensor was used. A medium to be chosen that had similar characteristics to vapor and a Lego sensor that could identify it. The medium choices were narrowed to light or temperature, as both have gradients similar to that of vapor, and both can be read by the Lego sensor. At the source these mediums have the strongest intensity and as distance from the source increases the gradients dissipate in strength. Light was selected because a light bulb is easier to install as a small compact source and is easier to view. Its ease of installation and view is useful when programming the robots. Thus, the mission of the robot team was to follow a light gradient to its source (the light bulb). The implications of this will be useful in the development of a similar robot team, equipped with artificial noses that find the source of any chemical (harmful or benevolent).


Journal of the RoboTeam’s Development
May 2003
The basic robot was designed and constructed. The initial robot design had three wheels, two in the back and one in the front, and two light sensors. The light sensors pointed up and were on opposite sides of the robot. This design allowed the robot to identify which side had a greater light intensity (right or left). All programming was done in Robolab, a pictorial programming language designed for children, but with many advanced capabilities.
The robot was programmed to base its movement on light intensity. The program compared the light intensity measured by the right and the left sensors. The robot then turned in the direction of the greatest light intensity. The first robot worked well and followed the light gradient to its source. A second three-wheeled robot was constructed, in addition to a large box with a light bulb inside.
The two robots were tested inside the box. It was determined that they lacked the ability to turn sharply enough to follow the light gradient. The placement of the light sensors also did not appear to be optimum. It was determined that the sensors should looking down at the gradient present on the floor or looking out for the direction of greatest light.

June 2003
A new robot design was developed and constructed. This robot’s light sensor looked outward toward the surroundings and rotated 180 degrees. The robot was programmed to compare light intensities in each direction. The new design eliminated the light sensor situated on each side of the robot. A new computer program was written for the robot and the robot was tested in the box. This robot design had a better turning radius, though still not sharply enough. The rotating light sensor seemed to be a good design.
The desired outcome is for the robot to turn sharply, but slowly, therefore, gearing up the wheels is not the best solution. Another motor was added to each side of the robot – yielding 4 motors powering the wheels and one micro-motor powering the rotating light sensor. The new design was tested and the new robot was then able to turn sharply and slowly in the direction of the most intense light.
Two more robots were built and programming for communication between two of the robots began. The RCX is equipped with an IR transmitter/receiver (similar to that in a TV remote control). For communication to occur between RCXs, there must be a direct line of sight between them. A direct line of sight is not always possible with the robot’s current design, as they are constantly moving and can be pointing in opposite directions when they need to communicate.
A robot design change was completed with the IR port facing up toward the ceiling, so that IR waves bounce off the ceiling of the box (or room) and return back to the other RCXs. The robots were programmed to send each other their greatest light intensity readings. The robot with the greater intensity reading would move and the other would continue to look, but remain stationary. Two robots with this programming were tested in the box. The ceiling of the box was lined with reflective white paper. Communication between the two robots failed to occur. To insure that the robots were attempting to communicate with each other, the IR ports were pointed directly at each other. The robots were then able to effectively communicate, indicating that communication was possible.
The emissivities of reflective, white paper were researched, in addition to other possible ceiling materials that could be effective. Research indicated that IR wavelengths reflect similarly, though not identically to visible light. Unlike visible light, IR is absorbed by the white surfaces and by mirrors. Aluminum foil was found to have a very low IR emissivity, and it was predicted that it should reflect the IR messages well.
The ceiling of the test box was lined with aluminum foil and the robots were then tested in the box. Once again they failed to communicate. It was hypothesized that the IR waves produced from the RCX were not strong enough and that stronger IR waves would allow for communication. The experiment was run again with IR ports turned up to high power. The robots successfully communicated.

July 2003
With IR communication working, the robots’ behaviors and the manner in which they process light intensity values was studied. Three robots were programmed to compare their individual light intensity value with the light intensity values of the two other robots. The robot with the greatest light intensity detection would become the master robot and continue to look and move in the direction of the greatest light. The two RCXs with the lesser light intensities detection would become the slave or non-active robots. These non-active two robots would continue to look for the direction of greatest light and their communicate light intensity value, but would not move in that direction.
A fourth robot was constructed, of the same design, and integrated into the current programming scheme.
Numerous test runs were conducted and it was noted that the robots often failed to register the target (the lamp) found when directly under it. The robots would pass under the target and circle around it repeatedly, failing to identify it. Another light sensor that pointed directly up was installed onto each robot. The robots were programmed to use this light sensor only to identify a set threshold light intensity. For the 100W light bulb in use, the threshold was set to 100%.
The rotation of the light sensor did not appear to be optimum. The period of rotation was measured with time, but it was difficult to match specific motor timings to 180 degrees. Each micro-motor varies in tension, strength, and battery power, thus each robot had to rotate the light sensor for a different period of time. Rotation sensors were then installed onto the four robots, to make the period of rotation of the light sensors constant. The rotation sensors performed with limited success. The sensors were measured at fast and slow speeds, and failed to measure 180 degrees of rotation accuratly. The time measured to move in 180 degrees varied on each rotation. A decision was made not to replace the rotation sensors, despite the measurement problems.

August 2003
The robots were programmed with distributed intelligence. The intent was to have each robot behave differently, depending on its relative light value. Each robot was programmed to place all of the other robots’ light intensity values into an ordered stack. Each robot then determined where its light intensity was in the stack and choose the appropriate motion based on its order.
Robot priority based behaviors:
1 (Greatest light intensity – Master) – Active Robot. Move in direction of greatest light and continue to look for direction of greatest light intensity.
2 (2nd greatest light intensity) – In-active Robot. Look for direction of greatest light.
3 (3rd greatest light intensity) –In-active Robot. Look for direction of greatest light.
4 (4th greatest light intensity) – Active Robot. Move in a random direction, right or left, and look to see if light intensity increases.
5 (least light intensity) – Active Robot. Move in a random direction, right or left, and look to see if light intensity increased.

The team of four robots was tested with the above scheme. The robots worked effectively to find the light source. Rotation sensor problems continued, but these were ignored for the time being and a fifth robot was constructed.
The five robot team was tested and found it to work as well as the team of four. However, the larger team overloaded the space available in the box. The robots were then tested in a larger space - a small room. The room had a high ceiling, but was otherwise of a workable testing area. When initially tested the robots were unable to communicate unless very close together. This indicated that the robots were receiving the IR messages diffusely when very close together, but they were not able to bounce IR waves off the ceiling. The white walls and ceiling were problematic as white absorbs IR waves.
The rotation sensor problem was tackled and different materials were researched to use for a false ceiling. The false ceiling would be lower and closer to the robots and constructed of a highly reflective material.
The rotation sensor were tested at varying micro-motor speeds and found to be more accurate at lower speeds. The sensors were still prone to turning much more or much less than 180 degrees and it was concluded that the sensors were not reliable for the duration of the experiment. Timing of the light sensor rotation was resumed.
Emergency blankets sold at REI are made of a thin plastic material and have a 99% reflectivity. Two different models of these blankets were purchased; a basic emergency camping blanket and an emergency camping tarp/blanket. The basic camping blanket consists of a sheet of very thin, very reflective plastic. The emergency blanket/tarp model is much more substantial. One side is red plastic “tarp” material and the other is highly reflective silver foil. The corners or the tarp have metal grommets that proved useful in attaching the tarp to the walls to the room.
The robots were tested under the false ceiling and their ability to communicate was maximized substantially with the new ceiling. The range over which the robots were able to communicate increased to nearly 8ft. The effectiveness of the robots’ programming was also tested to its limits. Since the robots could be placed further from the light source, they had to run longer to locate the light source. The longer run-time proved a good test for any errors in the programming and/or design.
Several small problems were fixed by minor adjustments and the project goals were achieved. A team of five robots were designed, constructed, and tested. The robots communicated with each other and used data to determine their course of movement. The robot team shared information in order to locate the target – a light bulb. Upon locating the target the remaining robots were signaled by the leader robot to shut off.
The summer project was completed; however, the robot team must be further developed, improved, and optimized. A comprehensive optimization plan for the team will be developed. This plan will test and determine optimum number of robots, optimum placement from the target, and placement positions on start-up.


System Limitations
Distance: the robots have been must be within 8 ft of each other to effectively communicate.

Target intensity: The target must have a known intensity as its source. For example the 100W lamp-shaded light bulb- has 100% intensity at its source. A smaller, 15W lamp-shaded light bulb- has only 90% intensity at its source, thus the detection threshold had to be modified.

Floor Surface: All floor surfaces yield a different response from the robots. The coefficient of friction of the floor compromises the ability of the robot wheels to turn. Highly waxed floors are “sticky” to the wheels and prevent sharp turning, while scuffed and very smooth floors have a low coefficient of friction and allow the wheels to turn easily.

Obstacles: The robots are highly susceptible to obstacles in their path. The obstacles may be the wall of the room or fellow robots. Due to the limited sensor ports available on the RCX, a touch sensor could not be installed to alert the robot when it has run into a wall, object, or another robot. This problem has been most pronounced with robot-to-robot crashes. As the robots come closer to the light source, room for movement is at a minimum and often results in a collision with another robot. Sometimes the robots manage to untangle themselves by turning in opposite directions, but often they have to be physically moved by the engineer observing the experiment. Additional sensor ports could be easily solve this problem.

Optimization of the Robot Team
Characteristics to Optimize
-Number of robots.
-Distance from target and distance around target.
-Position on start-up.
-Placement of light sensor (looking out vs. looking down)
System Improvements
-Rotation sensor control over light sensor rotation, rather than time
-Touch sensor to prevent obstacle limitation.
Experiment Chamber (room) Improvements.
-Black cloth to cover the white walls.
-Low coefficient of friction floor (coated plywood) with a numbered grid system to track robot movement.
-Improved, non-flammable lampshade.
-Timer.
Decisions
-Data recording system: Camera, hand diagrams or both.
-Recording of robot placement and direction on start-up.
-Grid system for robot floor.
-Data manipulation process to determine effectives of optimization parameters

Click here for more pictures>>

 

 

 

Pictures and Movies:


Click here to see the movie (requires Quicktime)

Click here to find more pictures>>

 

This material is based upon work supported by the National Science Foundation under Grant No. 0212046. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation (NSF).
Home Mission Program Projects Robotics People K12 Outreach Links
©2003 Tufts University