By Steve Meyer, Contributing Editor
Will your kids be begging you for this amazing robot in 2014?
Sometimes, a designer’s main mechatronic challenge is to make something that is normally big and complex into something small and inexpensive. Such is the case with MIP robots. MIP stands for Mobile Inverted Pendulum. It is one solution to the problem of mobility platforms that are expected to radically change our idea of what robots are and what they can do. Like the radio controlled car or R/C plane of the past, the addition of high-performance embedded controllers is transforming what mobile platforms can do.
In the community of robotic development, the University of California at San Diego (UCSD) is home to an outstanding group of innovators. Prof. Thomas Bewley has been teaching in the Department of Mechanical and Aerospace Engineering at UCSD for 15 years. In a 2003 final exam, Prof. Bewley challenged his students to work out the dynamics and control of a novel “robotic pogo stick” concept. The exam had a completely unanticipated side effect: several of the students approached Prof. Bewley after the course ended and went on to perfect the proposed system of locomotion, and the UCSD Coordinated Robots Lab was born. The Lab has been creating breakthrough mobile robot solutions ever since.
Among the many platforms that have come out of the lab in the last few years are Switchblade and IFling. Switchblade takes the classic treaded tank concept and adds a pivot at one end of each tread. These additional axes of articulation—and the associated control algorithms—create a mobility platform that can use its own body as a counterweight. This combination allows the unit to climb stairs and navigate obstacles much larger than is possible with other treaded platforms.
IFling is a simple two-wheel mobility platform that leverages a clever 3D-printed body shape to pick up and store ping pong balls simply by rolling over them, and then throw them on command with a jai-alai style arm. It is lightweight and nimble and practically breakdances on command. To see it in action, please visit the Engineering Exchange, www.engineeringexchange.com and search for iFling.
By mathematically modeling the mechanics of each type of platform and using model-based feedback control in a high-performance, low-cost ARM processor, the robot is able to both move and resist disturbances while executing complex maneuvers. In the case of MIP, the stabilization software combines the readings of a gyro, a two-axis accelerometer, and optical encoders on each motor to estimate the orientation of the robot at each instant. A control algorithm operating at 100 Hz then calculates the necessary voltages to send to two inexpensive brushed dc motors, which are connected to the wheels through a small 50:1 gearbox. The inexpensive ARM microcontroller used is quite fast, so there’s plenty of room for the other software, which shapes the vehicle’s unique personality.
Enter the toymaker
A few years ago, the management group at WowWee Toys, which has its U.S. offices near UCSD, was introduced to the UCSD Coordinated Robotics Lab by UCSD’s Technology Transfer Office. While the Lab’s proof-of-concept platforms were engaging, they were far too expensive to be the basis of a marketable toy.
Sensing the potential that might result from collaboration, UCSD and WowWee teamed up and began exploring new platforms with low-cost and design for manufacturability in mind. Timing was fortunate, as the semiconductor industry’s continually improving performance and cost brought a new ARM Cortex M0 processor at less than $1 at high volume. The cost of Micro-Electro-Mechanical-Systems (MEMS) sensors like accelerometers and gyros, which used to be prohibitively expensive, are now less than $1 each given the high-volume production of these sensors for smart phones. Add in toy-grade motors and gearboxes and low-energy Bluetooth modules, high performance, sophisticated vehicles become possible at the low price points demanded by the toy industry.
MIP is roughly 10% the height, 2% the mass, and 2% the price of a Segway, with similar dynamic behavior. MIP as an interactive two-wheeled mobility platform expected to come to market at under $100, with dynamic behavior comparable to a $5,700 Segway. Granted he’s just a little guy, but consider the transformation of technology the enables such a toy to run circles around his much bigger and more expensive cousin. Such a development could not be foreseen in 2002 when the Segway was first released and suggests the emergence of a range of new platforms enabled by this new cost/performance level.
MIP is fully interactive with its surroundings and is controllable by gesture or through Bluetooth link to any intelligent device. MIP has personality features, has a limited “voice” like R2D2, can dance to your iTunes and will be able to play games with humans. Given the capability of its computing platform, features for MIP are still evolving.
As part of the extreme manufacturability solutions needed to bring MIP to market, WowWee/UCSD figured out how to mount the motors, gearboxes, and custom encoders to one side of a printed circuit board (PCB), and the sensors and processor on the other side, maximizing the automated assembly and optimizing the cost of an otherwise mechanically complex system.
MIP’s first public showing will be at the Dallas Toy Fair (October 1-3, 2013), with broad availability planned for the first half of 2014.
The economics of robotics
The energy going into the development of MIP and related designs at UCSD is not simply for fun or profit. The field of robotics is evolving almost as quickly as the advanced smart phone technology that it leverages, and there is an acute awareness of the revolutionary new applications that are soon to be enabled by these advances. Unmanned mobile vehicles today are relatively big and expensive, and have been limited to applications in manufacturing, security and defense. Where there is a transformational change in cost and function, new applications in fire fighting, medical care and personal service will produce great change in our society.
There are two approaches to stabilizing MIP. Using Successive Loop Closure (see figure below), classical (single input, single output) techniques may be applied via two nested control loops—the inner loop regulating the body angle q(t) relatively quickly, and the outer loop regulating the horizontal vehicle position x(t) relatively slowly. Alternatively, state-space (multiple input, multiple output) techniques can be applied directly, thereby regulating both the body angle q(t) and the horizontal vehicle position x(t) simultaneously.
Education and MyMIP
To the people at Wowwee, “It’s not just about bringing technology to the toy industry, it’s about bringing technology to a broader audience.” Today’s children grow up with computers, the Internet, and smart phones as an integral part of their daily lives. This generation has technology as an integrated part of its perspective and they are bringing new ideas into the mainstream. A generation that grows up with embedded controllers and dynamic control systems as a part of their everyday experience will invent things we haven’t even imagined yet. Recognizing this the UCSD Coordinated Robotics Lab is developing a new undergraduate controls curriculum around home-built MIPs called MyMIPs.
MIP and many robotic toys that will follow are the direct result of the educational process. These platforms are not just compelling toys to inspire imagination and creativity, they also serves as a practical platform to teach science, mathematics, mechanical engineering, and embedded processing. Given the educational legacy it is fitting that the products of the UCSD/WowWee collaboration are being designed specifically with “hackability” in mind: A standard 3-pin UART connection to the onboard ARM microcontroller is being built into these toys. An open standard will be published for communicating with the ARM processor, allowing higher level functions to be generated by new platforms like the MyRIO board from National Instruments. It is easy to imagine more complex applications through a high level graphical language like LabVIEW, so that students can explore complex behaviors like visual recognition or collaborative play, like a nice game of robot soccer.
Teaching and research based on these platforms will move all the associated technologies forward, because the platforms are inexpensive but feature high performance. Potential applications are broad, ranging from control theory, embedded processing, adaptive communication networks, the study of swarming behavior, non-cooperative games, and collaborative Simultaneous Localization and Mapping (SLAM) algorithms using multiple vehicles to explore unknown and potentially hazardous environments.
UCSD Coordinated Robotics Lab