Using the Handy Cricket Robot Platform for Multi-Robot Research Rami Saikali Department of Computer Science Augsburg College Minneapolis, MN 55454 saikali@augsburg.edu Abstract The Handy Cricket is a small, mobile PIC microcontroller-based robotics platform that is versatile, cheap, and easy to deploy. Its small form factor, light weight, and low cost make it a very appealing platform to those who conduct multi-robot research. The feasibility of using such an inexpensive and minimalistic platform is explored in this paper. While it may be unrealistic to imagine the Handy Cricket itself being used in a real-life search and rescue operation, it provides a tangible platform for experiments in determining lower limits on communication and sensing abilities.
1 Introduction Complexity is an issue that plagues almost any field robotics included. When one wishes to accomplish a task using robots, one must ponder the best solution: a few, highly capable and expensive robots or several, somewhat limited, relatively inexpensive robots in a swarm deployment strategy. The latter, swarm robotics, is a relatively new approach to solving problems requiring robots. There are several advantages to multi-robot research as opposed to solitary robots. One of the most crucial is that disabled robots, or nodes, are not a crippling issue. In a singular robot environment, if the robot is either destroyed or disabled, the project must either end or another expensive robot must be deployed in its place. With swarm robotics the success of the overall task at hand does not depend on any one robot accomplishing the task. Additionally tasks are divided amongst several robots allowing more efficient use of time. Rather than searching spaces piecemeal one after another, the swarm robots would be able to push into and discover several areas simultaneously. A number of researchers are using swarms of small robots in their work. Dellaert et al.[2] developed a team of marsupials for search and rescue with small robots being transported by larger ones into key dispersion areas. Cheng, Cheng and Nagpal[1] have addressed formation control in robot swarms. Pearce et al.[6] experimented with the use of pheromones to control dispersion. Hsiang et al.[3] have developed a number of algorithms for the rapid dispersion of robots. Some of the research using robot swarms has been limited to simulation. The actual number of real robots used in trials has varied from a couple to a few dozen. The purpose of our research was primarily to discover and note the absolute minimum capabilities a robot is able to possess while still remaining useful in a search and rescue environment. A search and rescue environment is one in which the surrounding area to be explored is a completely destroyed environment. There are no, or very few, 90 angles in the destroyed environment. Visibility may be poor and there may be quite a few disabling obstacles. This type of environment beckons multi-robot research. To test the absolute bare minimum of allowable technology in a multi-robot research environment, we evaluated the feasibility of deploying the Handy Cricket robotics platform for multi-robot research. The Handy Cricket is extremely low cost, lightweight, and quite petite yet still very capable making it ideal for multi-robot research. 2 The Handy Cricket 2.1 Becoming Familiar With the Cricket The Handy Cricket board[5] is a PIC microcontroller-based robotics platform that aims to be very low cost while still providing numerous features. In essence, the board is a miniature computer with input/output and motor/sensor control. The Handy Cricket weighs only 0.2 pounds and occupies a space of only 3x3x0.5 inches. It is powered by four AA 1
batteries. On the main board there are two motor outputs, two sensor inputs, and two bus connection lines. Figure 1 shows the Handy Cricket board as viewed from above. Figure 1: The Handy Cricket board and all its connectors. (A) Motor Output[2], (B) Sensor Input[2], (C) Bus Connections [2], (D) Run/Stop button, (E) On/Off Switch, (F) Piezoelectric Speaker and (G) IR Communication. Photo courtesy of the Cricket website. The two motor outputs allow us to control two motors simultaneously, each capable of being operated both directions. The eyes of the Cricket board lie within its two sensor inputs. These two inputs may be connected to any combination of IR Reflection sensors, photocells, switches, or touch/pressure sensors. The sensor inputs of the Handy Cricket receive two types of input: Boolean (T-F) or a 0-255 range. Sensors like the IR Reflection sensor respond with a number between 0 and 255, which is the range from the sensor to the nearest object, while others like the photocell simply respond with a 0 or a 1 (a false or a true). The bus allows expanded functionality with the Handy Cricket platform. From the bus connection one may attach a 4-Digit LED number display, servo motor control interface, relay control interface, or one of two means of advanced motor/sensor control. Each of these expansion boards is a fraction of the size of the main Handy Cricket board and does not require additional external power. The two expansion boards with which experiments were conducted in the laboratory were the LED display and the motor/sensor expansion board. The motor/sensor expansion board provides two additional DC motor outputs capable of being run at eight separate power levels each and four analog sensor inputs. This allowed for the attachment of a total of six sensing devices using only one expansion board. 2
2.2 Programming the Handy Cricket The Handy Cricket uses a simplified version of Logo called Cricket Logo. There are several advantages to the ease with which one may write a program for the Handy Cricket. There are many built-in methods and routines and a getting started guide is included at the Handy Cricket s official website[5]. The board itself is programmed using an external communications board which connects to the host computer s serial port. All programming is done via IR communication with the Handy Cricket s on board IR transceiver. 2.3 Sensing and Communication One of the more crucial aspect of the Handy Cricket s performance in a research scenario is its communication and sensing technologies. Communication is paramount in a swarm environment in which multiple robots are being used. If the robots are unable to communicate with each other effectively the advantages of using multiple, cost-effective robots are essentially dissolved. All that remains in such a scenario are detached, relatively limited nodes. As shown in Figure 1, the Crickets have an IR transceiver affixed to the front of the board for communication. IR communication is low-bandwidth and near line-of-sight. However, even this limited communication ability is sufficient for multi robotics research. IR would be sufficient to communicate location data such as if a robot wishes to simply communicate the distance it senses to the next object, from its perspective, on its left, front, then right. Using an efficient data structure and expecting the appropriate type input this would take at most three bytes to communicate left, front, and right distances. This kind of one-way communication would take at most a few seconds to transmit. After this information is transmitted, the Cricket is free to go back to its previous duties. The three most popular sensors for the Handy Cricket are the photocell light sensor, the IR reflection sensor, and the micro switch touch sensor. Utilizing these three sensors as a means by which the Cricket can gauge its environment allows the robot to get an adequate picture of the world around itself. The most low-tech of the three sensors is the touch sensor; consisting of a micro switch, the touch sensor requires direct contact with an object before input is achieved. Because of the fact that the touch sensor requires direct contact the sensor becomes an excellent bumper sensor. By placing the micro switch on either (or both) the front or back end of the Handy Cricket the cricket is endowed with a last-ditch means of finding its way around. The touch sensor requires no special conditions to operate. Whether the environment is illuminated or dark, clear or smoke-filled the touch sensor is able to give the Cricket much needed tactile feedback. Due to its simplicity and effectiveness, the touch sensor was utilized heavily in the laboratory. The following Cricket Logo code demonstrates use of the touch sensor: 3
to loopy wait 10 a, on thisway b, on thatway waituntil [switcha] a, off b, off beep wait 10 a, on thatway b, on thisway wait 20 a, off b, off a, on thisway wait 10 a, off loopy end This procedure instructs the robot to move forward until input is gained from switch A. After input is gained from switch A (contact is made with a wall or some other object) the robot shuts its motors off, reverses direction for two seconds in order to back up, then enables one motor for a second to turn it in another direction in an attempt to move around the obstacle. The second, more high-tech, sensor of interest is the photocell light sensor. The photocell light sensor acts as a switch that remains off until light is introduced to its surface. By shrouding the photocell light sensor in a piece of cardboard or paper tubing several inches long, ambient light is attenuated from the photocell s input. This method allows the photocell to become more of a directional implement. By this method the robot could be instructed to search for a light source or continue a wandering procedure until a light source is found. The last, and arguably the most interesting, sensor is the infra-red reflection sensor. Rather than a simple on/off state, the IR reflection sensor reports back the distance from a solid object that renders a reflection of infrared energy. In the Cricket s case the number is between 0 and 255. Using this sensor gives a rough estimate of the Cricket s distance from its surrounding environment and aids in obstacle avoidance. 3 Designing and Navigating Destroyed Environments The purpose of our research was to test a tree-based dispersion algorithm using multiple simplified robots in a search and rescue type scenario. In order to test the usefulness of such a configuration we designed a destroyed environment. 4
Additionally a proper means of supporting the Handy Cricket board itself and making the board mobile had to be implemented. 3.1 Designing the Destroyed Environment The environment was first designed in the Player/Stage platform by Lava K.C.[4]. The destroyed environment that was constructed was modeled after his simulated environment. Figure 2 shows this environment. Figure 2: The destroyed environment in Player/Stage. A destroyed environment was designed using foam-core board. Development of the environment was facilitated by the modularity of the pieces of board and the ease of their connections. Figure 3 shows the Handy Crickets navigating in this environment. The most important aspect in the design of the destroyed environment was that there would not be simple 90 corners to navigate. One of the difficulties of navigating a destroyed environment is that normal room structure is not preserved. Navigating perfectly rectangular rooms is quite easy, however navigating a room with several sharp corners of undetermined angles is much more difficult. 5
Figure 3: The Handy Crickets navigating through an actual destroyed environment. 3.2 Designing a Platform for Deployment To navigate the environment the Handy Cricket boards were attached to Lego bodies and motors. Connecting the Handy Cricket boards to a Lego platform allowed quick attachment and removal of sensors and motors. A fully assembled Cricket on our lego platform is shown in Figure 4. 4 Experiments on Sensing and Communication Since Cricket processing power is quite low, analysis of input from sensors must be minimal. Additionally, because the the Boolean input from the touch and light sensors are minimal, we focused our efforts on testing the capability of the more sophisticated IR reflection sensor. The IR reflection sensor seemed to be slightly inconsistent in its accuracy during our tests. In actual tests in the lab it was found to be a little rough in the results it reported. Past approximately one foot the sensor reported a maximum value of 255 or its maximum distance. When the sensor was brought near a wall the relation to the reported value was not linear and we could ascertain no accurate relation between distance and reported value. Rather, the IR reflection sensor reports values ranging from 0 to 100 when the cricket 6
Figure 4: The Handy Cricket fully assembled with its sensors on our Lego platform is closer to an obstacle and values above 200 when it is much farther. This allows for approximate distancing and obstacle avoidance. In terms of communication range, IR, however, proved quite viable. In direct line of sight 100% receipt of signal was achieved at a distance of over 11.5 feet. Even at an angle of 45 degrees to either side the signal integrity remained at over 10 feet. To measure signal intensity one Cricket was programmed with transmitting code and the other was programmed with receiving code. When one Cricket would transmit a value of 1 via the IR port it would also emit a beep. It would do this every two seconds. The other robot would wait until it received a 1 via its IR port and then it would emit a higher pitched beep. In this manner we knew that the other robot was receiving the signal if it was reporting back with a high pitched beep. To ensure the signal was received during testing, we waited to listen for 30 transmission beeps and 30 reception beeps before we declared that transmission was accepted at 11.5 feet. As distance between the crickets increased beyond 11.5 feet (138 inches) the signal strength dropped off by 10% (loss of 3 beeps for every 30) every two inches the receiving Cricket was moved back. In an attempt to determine what factor lead to the surprisingly long distance for the IR communication, black fabric was purchased to cover the walls during the ranging of the IR sensors. The concern was raised as to whether the white walls of the laboratory were reflecting the IR energy and giving a deceptively long range to the sensors. It was found that draping the adjacent wall in black fabric had no significant bearing on the distance the IR transmissions were received. While IR seems to be a limited communications method at first, the low power consumption of an IR transceiver, its excellent range, and tolerable bandwidth have made it an excellent choice for the Handy Cricket platform. 5 Conclusion Over the course of our research in the laboratory we have become quite familiar with the Handy Cricket in an attempt to ascertain its capabilities. We have attempted to determine 7
whether or not the Handy Cricket would be an appropriate platform to implement the algorithms we have developed in simulation for search and rescue. The platform does have its bonuses; its low weight and small size make it quite desirable. Its sensors are quite usable for navigation. We did find, however, that the IR reflection sensor seemed to fall short when used to detect exact distances. On the other hand IR communication proved remarkable in its range and usefulness. The programming language, Cricket Logo, while still not as full-featured as other programming languages may still be all that is neccessary to handle the bit strings that our simulated robots use for cmmunication. Other tasks that yet need to be explored include implementing the dispersion algorithm developed in simulation[4] on the Crickets, testing inter-robot communication, and how to effectively storing the information needed by each robot. 6 Acknowledgements This project was supported by National Science Foundation grant EEC-0538740. Thanks also to Lava K.C. for input and advice on this project. References [1] CHENG, J., CHENG, W., AND NAGPAL, R. Robust and self-repairing formation control for swarms of mobile agents. In Proceedings of the National Conference on Artificial Intelligence (July 2005), AAAI. [2] DELLAERT, F., BALCH, T. R., KAESS, M., RAVICHANDRAN, R., ALEGRE, F., BERHAULT, M., MCGUIRE, R., MERRILL, E., MOSHKINA, L. V., AND WALKER, D. The georgia tech yellow jackets: A marsupial team for urban search and rescue. In AAAI Mobile Robot Competition (2002), pp. 44 49. [3] HSIANG, T., ARKIN, E., BENDER, M., FEKETE, S., AND MITCHELL, J. Algorithms for rapidly dispersing robot swarms in unknown environments, 2002. In 5th International Workshop on Algorithmic Foundations of Robotics. [4] LAVA K.C. An Algorithm for Dispersion of Search and Rescue Robots. In Proceedings of the Midwest Instruction and Computing Symposium (MICS 07) (2007). [5] MARTIN, F. The Handy Cricket. http://www.handyboard.com/cricket. [6] PEARCE, J. L., POWERS, B., HESS, C., RYBSKI, P. L., STOETER, S. A., AND PA- PANIKOLOPOULOS, N. Dispersion of a team of surveillance and reconnaissance robots based on repellent pheromones. In Proceedings of the 11th Mediterranean Conference on Control and Automation (June 2003), MED 03. 8