There is a lack of low-cost robots available which satisfy the functional requirements of an educational environment. In order to satisfy the pedagogical goals outlined in this project, we designed and built custom hardware to support educational robotics.
Initial Prototype
Our primary objective for the hardware development portion of the project was a low price-point for the final product. To achieve this goal, we had to investigate which features were necessary in an educational robot and aim for a minimum feature set in the final design. We started with a low-cost off-the-shelf robot and incrementally added features as necessary, based on feedback from student trials.
The base for our prototype development was a commercially available scribbler robot manufactured by Parallax, Inc. The Scribbler is a two-wheel skid-steer design with a two-tone speaker, some infrared and photo-resistor sensors, and not much else. We knew an educational robot needed to support dynamic student interaction, so our first addition was a wireless Bluetooth RF link. We wrote a simple software server that runs on the scribbler and allows the PC to wirelessly control the motors and speaker and query the sensors. The goal was to allow students to write, compile and execute programs on their PCs instead of bothering with the native programming environment of the Scribbler’s internal BASIC Stamp microcontroller. We specifically chose a Bluetooth radio because Bluetooth is widely available, inexpensive, and performs well in a multi-transmitter classroom scenario.
Figure 1 – Initial Prototype using off-the-shelf components.
This prototype solution worked well, and we used the platform extensively in workshops and classroom situations. However, two problems were evident from our experimentation. First, the off-the-shelf Bluetooth radio we were using was too expensive (more than the robot itself!). Second, the robot was a bit sensor-deprived. Specifically, it did not have a camera, a feature that was frequently requested by students and teachers in our trials. The inclusion of a camera makes available significant educational opportunities in the fields of image processing and computer vision.
The Fluke Robot Upgrade Module
Having learned from our prototype robot, we were ready to move another step closer towards the final solution. We decided not to discard the low-cost scribbler robot base. Instead, we aimed to reduce costs and add necessary features by developing our own custom circuit board. The primary elements of the circuit board would be a Bluetooth wireless radio and a camera. Additionally, this circuit board would have to seamlessly integrate with the scribbler robot base.
Figure 2 - IPRE fluke circuit board installed on a scribbler robot.
Operation of ourfluke robot upgrade module is completely automatic. Once attached to the scribbler robot the module uploads the Myrosoftware server into the scribbler’s internal program memory. The module then instantly provides a wireless link between the student’s computer and the scribbler robot.
We chose to implement a chip-level Bluetooth radio rather than integrate a pre-engineered module. By buying the Bluetooth chips directly from the manufacturer, we were able to significantly cut costs. However, it took much development effort and many board revisions to produce a reliable RF printed circuit. The Bluetooth radio circuit operates at 2.4GHz, which is such a high frequency that standard circuit board layout assumptions no longer apply. For example, a circuit board trace that connects two components at 2.4GHz must be viewed as a parasitic indictor and capacitor instead of a simple wire. It is difficult to predict in advance the exact value of these parasitic components so an iterative process is necessary to optimize the circuit. The penalty for a non-optimum circuit layout is poor radiation efficiency and increased levels of harmonic frequency interference.
For the Bluetooth chip itself, we chose to use the Bluecore4-external product manufactured by Cambridge Silicon Radio. The main advantage of these chips is that they allow the Bluetooth firmware to be stored in external rewritable flash memory, which avoids the expensive mask charges associated with on-chip ROM memory.
In addition to the Bluetooth RF link,the final design included a camera, infrared proximitysensors, battery voltage monitoring, and a bright LED. Under the hood, a 32-bit ARM LP2106 microcontroller manufactured by Phillips allowed for demanding software algorithms such as color segmentation and JPEG image compression. The infrared proximity sensors were added to the design to expand the scribbler’s existing bump sensor functionality and allow the robot to avoid obstacles while driving in reverse. The infrared sensors also provide for an alternate method of robot-to-robot communication. The battery voltage monitoring is necessary to gracefully shut the system down before low battery level causes erratic robot behavior. The bright LED can be tracked by the camera on another robot and is useful for multi-robot interaction.
Mass Production
We were ready to distribute our fluke upgrade module to students at Georgia Tech, Bryn Mawr and early adopters at other schools across the country. However, two significant hurdles to distribution of our hardware remained: FCC certification and mass production.
The FCC legally requires that all electronic hardware sold in the United States passes an inspection to verify the device does not violate FCC RF radiation guidelines. Our custom Bluetooth radio circuit would have to be tested by an independent lab. The guidelines are strict so this certification process was difficult. It took several more board revisions, but we eventually passed. Our official FCC certification ID is VPU-4970726505.
The FCC is primarily interested with limiting radio transmissions to legal frequencies. Portions of the frequency spectrum are allocated to critical national infrastructure projects such as aeronavigation, and consumer products cannot interfere with these systems. The next figureshows 5 different versions of the fluke circuit board. Each version reduced the radiated harmonic signal level and brought us closer to FCC certification. The primary change was to the length and thickness of interconnecting copper traces, as these traces appear as parasitic circuit elements in a high frequency design.
Figure 5 – Five revisions of the board layout were necessary to pass FCC certification testing.
In order to make thousands of our circuit boards, we would have to optimize the design for manufacture, carefully choose available components, and establish a supply chain for each of the almost 200 components on the board. Especially difficult was the camera chip, which we had to buy surplus from Hong Kong because it had been discontinued in the time between our prototype and final circuit board.
Over the last two years, we have had 2300 of our boards produced and nearly all of them sold and distributed to classrooms and educators across the country. We have another production run of 1000 robots slated for manufacture in mid-2009.
The Gyro robot
We have deployed thousands of our custom-built first- and second-generation educational robots in classrooms across the country. Using lessons learned from this experience, we have designed a new robot, our third-generation educational robot product, that builds upon the strengths of the previous platforms and addresses problems discovered during deployment of the previous hardware in realistic classroom scenarios. Using our experience in mass production, we have eliminated the Scribbler robot base and incorporated its functionality with our fluke product, resulting in our Gyro robot design.
We intend to solve the worst problems with the scribbler robot base. The number one student complaint we’ve heard from our classroom trials is that the current robot does not drive perfectly straight. This problem can be traced back to the poor quality of the scribbler’s motors. Students also don’t like that they have to buy so many batteries for the current solution. Our Gyro robot prototype incorporates an integral rechargeable battery cell.
The electronic and mechanical aspects of the Gyro robot have been designed and simulated to verify the design is valid. A physical prototype is still under development. One significant addition in the Gyro design is a large LED scrolling sign, which we believe will increase the robot’s expressivity and encourage more student-robot interaction. Another important addition is a microphone, which will allow students to record and playback sounds with their robots. A CPLD video buffer has been added to increase the frame rate and resolution of the onboard camera. We have also optimized the electronic circuits in the robot for mass production by eliminating many difficult-to-procure IC components and replacing them with simple commodity transistors, resistors and capacitors. The next step in this project is to finalize the physical gyro robot prototype and begin student trials.