"We have been developing robots as the interface between a human and the surrounding environment. This time, we made Papero compact enough so that it can be carried around. We are going to take Papero out of the laboratory to pursue better communications capability," said Tsutomu Temma, general manager of NEC's Incubation Center.

After it brings the robot out of the lab, NEC hopes the public will one day welcome Papero into their homes as a mobile terminal that can carry out home-security functions; help care for the elderly, the young and the infirm; and provide a superior communications gateway.

The 5-kg, 15-inch robot sports three wheels, a jointed head, a built-in wireless modem and the capability to perform seven kinds of recognition/sensing tasks. Two 330,000-pixel image sensors allow Papero to recognize human faces and obstacles. The current lab model can successfully identify about 10 people, said Yoshihiro Fujita, project manager at the Incubation Center. Papero's image sensors scan for human faces. Once it finds a face, the robot identifies the scanned image by cutting out the central portion of the face and abstracting the characteristics. The image sensor also functions as a camera to record video messages.

Sound sensing employs four microphones. Three are used to detect the direction of the sound source. By measuring and calculating the time differences among the microphones' reception of the sounds, Papero can pinpoint the location of the sound source and turn in the appropriate direction.

The fourth microphone is used for voice recognition. Papero understands about 650 words and can utter about 3,000 words at present, but Fujita said there is no "technical limitation" on the number.

"Through the Internet, new scenarios can be downloaded to the robot to increase the number of words," he said.

Papero employs two types of sensors to detect physical contact. Sensors trigger when someone pats the robot on the head or strokes its head, and Papero will react to different family members' pats and strokes with appropriate verbal responses. A lift sensor, meanwhile, detects when Papero is lifted and carried.

For successful navigation around obstacles, a floor sensor and five ultrasonic sensors respectively detect bumps on the floor and measure the distance to objects. CCD sensors help the robot steer around obstacles.

Armed with its array of sensors, the autonomous robot can successfully move about a room, recognize a family member and carry on a simple conversation. The user might instruct the robot to perform a certain task, such as dance, play a riddle game or remotely operate TVs and other electrical appliances in the home. The robot may respond with personal information for the user, such as the time of day or personal messages.

"Papero will grow to become a home robot terminal or mobile robot terminal, which has not existed before," said Fujita.

Those Jetsons-age functions are enabled by relatively mature, established technologies. "The hardware architecture is based on that of notebook PCs," said Fujita. Papero is powered by a 500-MHz Celeron processor and the Windows 98 operating system. The CCD image sensors are connected via an IEEE-1394 interface; the other sensors are connected through USB.

"Since the OS platform is Windows, R&D is easy not only for us but for other people. We are going to maintain this PC-like environment for the time being to conduct R&D," said Fujita. To simplify programming of the robot's actions, dialog and behaviors, for example, the researchers developed the Action editor and Scenario editor, both of which run on Windows.

But Papero's current grounding in the PC environment "does not mean the platform is fixed to a PC base," Fujita said. When Papero is readied for release as a product, "we can change the CPU and OS if necessary."

NEC researchers have built about 30 units of Papero for distribution to outside development collaborators. The company intends to conduct further study with various universities on recognition technology, autonomous technologies and the robot's interactions with humans.

NEC will collaborate with third-party companies on practical applications for the robot. As for when models might hit the market, Temma said only that the "the robot is still under study at laboratories."