The main difference in price most likely has to do with the different sensors and mapping techniques that the robots implement. Truth be told the Roomba is really a "dumb" robot which only reacts to situations (i.e. when it bumps into something or detects a IR beam). It will then just randomly turn in another direction. Now given enough time, it will eventually travel over and clean a portion of the room it is in, but it will probably not get everything (given an infinite amount of time it should, but it doesn't have an infinite battery life to do this). This type of design is very simple and easy to make. Bump and IR sensors cost less then $1, and there really is not much processing required, just a simple random number generator to pick a number against a table of different commands (each command most likely being turn right for x amount of time, or turn left for x amount of time). This type of processing can also be created very cheaply, probably under $10 in parts and construction. So in the Roomba's case you are paying for a $180 cord-less vacuum with about $20 worth of "smarts" to it.

The Trilobite is using a version of radar to actually create a map of its environment. It then uses this map to plot where it currently is located and where it has been in the past. It can use this information to figure out a way to travel efficiently around the entire room and cover all the places that it can get to given its size limitations. I've designed similar robots myself, the purpose of those robots was to be able to tell the robot a destination and have it figure out a way to get there without running into obstacles in its way. This type of system requires a processor and memory to handle the style of programming needed to do these functions. A real-time planning and event based operating programming is used to instruct the robot what to do. It first will take sensor sweeps with its radar/sonar and create a map of the area around it. This map is used to mark obstacles that it detects with the sensors as well as mark where the robot is currently located and where it

was in the past. Rudimentary versions can be built using a processor, memory, and the sensors. In addition to having the bump sensors which I have discussed above, the sonic sensors are a little more complicated. A cheap version can be done by using Polaroid sonar distance sensors (these where created by Polaroid for their camera's to help in auto-focusing by calculating a more precise distance to the target). Even these sensors are not very cheap (around $30). And they have an audible clicking sound as the frequency of the wave is within the human hearing range. The Trilobite is using sensors outside the audible spectrum which will also allow them to be more accurate as background interference will be reduced because it is less likely that there will be other devices creating waves in that same frequency range as the device. This type of mapping will also need to handle "echoes" (reception of the burst from another location at a longer delay then the initial reflection of a

n object would create, thus making it believe that there is an object farther away in the distance when there may be nothing there at all). Echoes are not the only obstacle that needs to be addressed. Room reflections, corner blindness conditions, and close wall blindness are all other obstacles that need to be accounted for.

Take this situation, the sonar/radar is aimed at the corner of a box. The reflection from the box will not be detected because the beam will reflect off the box at a side angle and never return to the robot. The robot would never see the box and thus be blind to it (this is part of the reason why the stealth fighter has such a small radar image, because of all the angles on the plane, the radar beam is reflected away from the plane, but not back at the radar source, thus the radar does not detect it).

Creating a program to properly handle all of these problems correctly is very time consuming and difficult (if it was simple, we would all own cars that drive themselves). It is not 100% correct all the time and it will still have some problems detecting objects with sonar/radar, but the Trilobite will be able to know that it has traveled across all the areas it needs to clean.

The full processing power and sensor network needed to create a decent mapping system will cost around $80 with just one sonar/radar sensor. With more sensors, the price will jump about $30 per sensor until it reaches a point where a more powerful processor is required to interpret all the data being sent by the sensors. I do not have the exact specs on the Trilobite, but let us assume it uses between 2-4 sensors so that will cost around $120-200 in parts alone. Factor in the cost for programmers to develop the software to run the robot and there is probably another $200 per robot in cost. So you are looking at about 20x in price of smarts put into the Trilobite as there are in the Roomba.

Does this justify the almost 20x total cost? I don't know, as I don't own either. But at least it gives the reason for the price difference. Would you rather own a robot which will most likely be able to completely sweep your room and be able to return to its docking station, or a robot which may or may not be able to clean your room, but will probably get to a large portion of it before running out of power? Now that is the question that you need to answer.