SAN JOSE, Calif. Concealing very important thinking beneath the veneer of a humorous lunchtime speech at the Custom Integrated Circuits Conference here on Tuesday (Sept. 23), Sony Corp. corporate advisor Tsugio Makimoto suggested there is a next big thing waiting in the shadows to rescue the IC industry: a robot.
Makimoto's projections have to be taken seriously, as he was the founder of the observation since given his name that demand and capacity in the semiconductor industry were cyclical. Hence, an audience of chip designers paid close attention not just to his entertaining descriptions of recent Sony entertainment robots and to his breakdown of critical IC content in the devices, but to his views on what role robotics would play in the future of the industry.
Makimoto paid tribute to the Czech playwright who coined the term robot and to science fiction writer Isaac Asimov, who created the fictional but still influential three laws of robotics. He then described the four-phase chronology of robotics.
In the first phase, he said, robots were playback devices that performed prerecorded routines. In the second phase, robots became sensor-driven devices that interpreted input from a limited range of sensors to determine their next move in a well-defined environment.
The third phase, Makimoto said, was the entry of robots into unstructured environments where they required sufficient local intelligence to interpret the sensor data without the constraints of a predefined world.
Finally, the researcher claimed, robots have recently moved beyond just the ability to navigate unstructured environments to the ability to coexist with humans. Makimoto made clear with video clips that for him, coexistence meant interacting with humans on a verbal and even an emotional level.
The Sony advisor described the IC content of two well-known entertainment robots, the AIBO pet and the SDR, or Sony Dream Robot. AIBO, the dog-like device, is driven by a single 64-bit RISC CPU with 32 megaytes of memory, Makimoto said. Dedicated interface chips serve an array of sensors, motors and actuators that articulate the device and give it vision and speech.
But the SDR, a 6-kilogram human-like robot, is a device of another order, he said. SDR, which has impressive abilities at image recognition, bipedal motion, grasping and articulating its limbs, relies on a central processing cluster of three 64-bit RISC chips and a total of 192 megabytes of DRAM. The computing cluster is supported by 29 16-bit microcontrollers, 23 DSP chips, four ASICs, three FPGAs and 16 megaytes of flaish memory.
Most of the resources serve SDR's array of sensors, actuators and motors. The most recent version of the device has 38 joints and 67 sensors, including an array of thermal sensors and pinch detectors. Vision comes from two color CCD cameras. Hearing, which can localize sounds accurately while providing speech recognition, comes from six microphones.
Makimoto used this illustration to make two significant points for the semiconductor industry. First, he said, for ICs in devices like SDR, the Moore's Law treadmill was irrelevant. He added that most of the robot devices depend more on cleverness of design than on process advances. He cited in particular the vision chip that combines a pixel array with an array of analog image storage cells and comparators along with a set of accelerometers based on surface Mems integrated with electronics.
The role of Moore's Law, he suggested, was to provide the continual increase in computing power necessary to take us from today's embedded processors to the estimated 108 Mips necessary for human-level intelligence. Makimoto outlined a set of challenges, the most demanding of which was the RoboCup: the drive to field a team of robots that can beat a championship human soccer team by 2050.
Makimoto also presented market estimates showing that "personal" robots for entertainment, to work in tandem with humans, to serve them or as companions could become a larger market than industrial robots in a few years, reaching perhaps $20 billion by 2010.
By then, robots for mostly unforeseen uses would become important as well, he said. Hence, robotics promoises to be the next wave of massive investment and consumption for the electronics industry, following the successive waves of analog electronics, PCs and digital consumer and networking products, Makimoto argued. In effect, the IC industry's next big thing.