Building kilobots and revising kilobot design for improving the optical response

WCU Author/Contributor (non-WCU co-authors, if there are any, appear on document)
Anik Tahabilder (Creator)
Western Carolina University (WCU )
Web Site:
Yanjun Yan

Abstract: Inspired by the emergent behavior of swarms, we want to eventually use a distributed self-organizing swarm of robots for shape formation. To verify the idea using real robots in the experiment, we need to first build more Kilobots to enlarge our repository of Kilobots. Kilobot is a kind of small robot with a 33-mm diameter that was originally designed by Harvard in 2012 and redesigned at WCU with the simplified building process in 2016 (WCU Kilobot version 1.1). Based on the earlier design, we have redesigned the Kilobots further (with three revisions in version 1.2, 1.3, and 1.4). This research work describes the challenges and solutions in building and debugging Kilobots, as well as the planned shape formation operation. Kilobots are built in-house using reflow soldering for surface mount components and hand soldering for through-hole components. A systematic debugging procedure, as well as the most commonly seen issues and their solutions, are described based on our building and testing experience. The WCU Kilobot version 1.1 was designed in PADS, and yet we no longer had the license in PADS. Therefore, we redid the schematics and PCB layout in Altium Designer, and enlarged the spacing between the crowded components, in WCU Kilobot version 1.2. Although the design of version 1.2 was nearly the same as in version 1.1 with only added spacing, it was redone in Altium Designer that we could continue to maintain a license, and hence our later revisions were possible. In shape formation, a phototaxis movement (moving away from light) is the driving force in the large-scale reductive approach, and yet the original Kilobot design allows such movement only in a dark room because of the ambient light sensor output is saturated at a low illumination level. An experiment was conducted to examine the saturation of sensor reading at increasing lux levels with different phototransistor’s emitter resistances, and a new resistance value of emitter resistance was proposed and implemented in our Kilobots (version 1.3), to ease the experiment lighting condition, making it more lenient and convenient than before, even at daylight. An earlier capstone experiment in 2018-2019 seemed to indicate that the flash memory ofATmega328P, the microcontroller on the Kilobot, was not enough to handle the calculation when more than three Kilobots with known or calculated locations were used for multilateration-based locationing for the next robot that needed to calculate its location. To address this issue, we have updated the design of Kilobot to replace its ATmega328p (with 32 Kbytes memory) microcontroller with ATmega1284, which has 128 Kbytes of flash memory for programming(version 1.4). In addition, we also inspected the feasibility of installing two ambient light sensors at opposite sides of the Kilobot and found the version 1.3 was more sensitive than version 1.1 to provide distinctive readings even at a distance increment of one Kilobot diameter, which meant that the Kilobot could easily tell the direction of the light with two sensors. Given the new microcontroller in version 1.4 with more IO channels, we further revised it to add a second ambient light sensor, which will help to give us more control on the Kilobot when they perform a light based movement, such as in shape formation.

Additional Information

Language: English
Date: 2020
building, debugging, Kilobot, robotics, shape-formation, swarm
Swarm intelligence
Robots -- Design and construction
Debugging in computer science

Email this document to