A Low-Cost, High-Performance Robotics Platform for Education and Research

The authors have developed a new robotics platform for researchers and educators, providing an improvement in processing power compared to existing low-cost solutions while maintaining a low cost. The platform includes a microcontroller board, GNU-based development tools, and a suite of device drivers and basic utilities. The authors successfully used the platform in a one-month long robotics crash-course at MIT. Motivation for a New Robotics Platform While several expensive robot kits are available commercially, many researchers cannot afford to spend thousands of dollars on a basic platform. Educators, attempting to improve the robot to student ratio, are even more constrained. Several robotics platforms have emerged that are affordable for both, including Lego Mindstorms and the Handy Board. However, for researchers and educators hoping to tackle harder robotics problems, the features of these boards may not be sufficient. We have developed a robotics platform that is comparable in cost and easeof-use to the Handy Board yet provides greater computational power and additional features. Existing platforms have many good traits that are worth emulating. The Handy Board serves as an example of a successful platform that enjoyed a long lifespan. Some of the traits that we believe are responsible for its popularity are simple operation, simple programming environment based on a commonly understood language (C), straightforward electrical interfaces enabling “hardware hacking,” and a general immunity to abuse, making it appropriate for use by inexperienced students. The primary disadvantage of the Handy Board, Lego Mindstorms, and other similar platforms is their limited computational power. For example, the Handy Board, able to execute only about 500k instructions per second, cannot handle video processing— certainly not motion flow. The 32k memory capacity is also a limiting factor; a single 320x200 video frame cannot be held in memory at once. Robots building maps of their environment can also quickly exhaust the available memory. Another controller board deserving mention is the Compaq Skiff. Based on the StrongArm SA-110 microprocessor, the Skiff has vast processing power (up to 233MHz) and can address 16 megabytes of DRAM. It also has a large number of I/O ports for interfacing to various types of sensors. The Skiff, however, is an extremely complicated board, consisting of two stacked PCBs with high-density parts on both sides of each PCB. The primary bus interface is PCI, which makes “hardware hacking” very difficult for novices. Lastly, the cost of the Skiff is likely to be a hardship for many. The Compaq Skiff can be programmed using the GNU toolchain, which provides a performance improvement over the HandyBoard’s Interactive C, which is interpreted. It also enables users to use existing and highly-optimized libc and mathematics libraries and the powerful gdb debugging suite. An additional shortcoming of many existing boards is the lack of explicit support for sophisticated peripherals. It was our goal to provide support for high-current (2A per channel) motor drivers and dedicated inputs for quadrature phase optical encoders, for example. While these capabilities can usually be added to existing boards, it often requires specific knowledge of the internals of the microcontroller and the board’s other components. A New Microcontroller Board After developing several small robots, and with plans of running an ambitious robotics course, we were motivated to design a new microcontroller board that could serve the role that the Handy Board did, while providing the computational power, memory capacity, and other features that we wanted. Figure 1. Block diagram of controller board Hitachi SH-2 2MB DRAM CPLD Analog Mux Switches (DIP and momentary) Digital IO 16 Analog IO