Sony Corporation today announced the development of a prototype small biped entertainment robot “SDR-4X” that can adapt its performance to its environment and situations found in the home to further develop the possibility for a biped-walking robot. Richer communication with people can be realized by a variety of sensor systems, performance control software based on memory and learning, and a flexible biped walking mechanism.
While improving the output performance of small actuators that drive each joint, a newly developed “Real-time Integrated Adaptive Control System” controls a total of 38 joints in the robot’s body in real time, based on information from various sensors gathered in real time. More advanced movements are realized by enabling biped walk on irregular and tiled surface plus posture retention control under external pressure. In addition, stable and flexible walk can be achieved by the real time production of walking patterns such as pace and rotation angle in accordance with various situations.
Using two CCD color cameras for image recognition, the SDR-4X can detect the distance between itself and an object by processing the parallax of the two cameras. These cameras allow the robot to perceive the figuration of an object and range between itself and the object to automatically produce a route in order to make its way around the object.
In addition to image recognition, sound recognition, and sound synthesis technologies, communication and movement control technology based on memory is incorporated in the SDR-4X to further enrich communications with people. The SDR-4X can recognize an individual person by detecting the front facial image captured by a color camera. The robot can also detect the direction of a sound source and recognizes an individual speaking by utilizing 7 microphones located inside its head. By using embedded wireless LAN communication functions, the robot can synchronize data processing with an externally connected PC, which enables continuous speech recognition of many vocabularies.
Information on people and the location of an object obtained through image recognition technology is used as short and long term memory information. Based on this information, the SDR-4X can realize more complicated communication and movements. By inputting music and lyric data into the robot, it can produce a singing voice with vibratos and the composition of emotional, dynamic singing through voice synthesis can be realized to improve the robot’s entertainment quality.
The SDR-4X will be shown at ROBODEX 2002, an exhibition of “Robots as Partners”, from Thursday March 28 to March 31 for 4 days at Pacifico Yokohama in the Yokohama Minatomirai area.
Prototype “SDR-4X” Key Characteristics
1) Real-time Integrated Adaptive Motion Control
Integrated Adaptive Control can be realized in real time for walking on irregular and/or tilted surfaces and retaining posture under external pressure (i.e. when pushed).
– Ability to walk on an irregular surface up to 10mm approximately in variation
– Ability to walk up and down on a tilted surface up to 10 degrees approximately
– To prevent falling over under external pressure, the SDR-4X automatically performs footfall and step back to maintain its standing posture.
– If it falls over, damage will be limited by the robot’s flexible joint control and effective damage avoidance posture retention.
2) Real-time Gait Pattern Generation Control
This technology produces the necessary walking patterns by altering walking pace, cycle and rotating angle in accordance with the situation and environment, based on information from the robot’s sensors. This achieves a stable and smooth gait for autonomous walking.
3) Real-world Space Perception Technology
With 2 CCD cameras embedded in its head, the SDR-4X can detect the distance between itself and an object and perceive the range in-between and the figuration of an object. Based on this information, SDR-4X can automatically calculate a route to make its way around the object. 7 microphones equipped in the head make it possible for the SDR-4X to detect the direction of a sound source.
4) Multi-modal Human Interaction Technology
1. Individual person detection, recognition learning technology
– Ability to detect and recognize a face (front face) against a complex background
– Ability to memorize up to 10 individual faces through the learning functions
– Ability to recognize individuals by the tone of their voice.
2. Continuous speech recognition and unknown vocabulary acquisition
– Ability to recognize continuous speech for many vocabularies by synchronizing data processing with an externally connected PC via wireless LAN
– Ability to learn and memorize new words not listed on its dictionary.
3. Conversation, performance control technology based on its short-term and long-term memory
In addition to short-term memory functions to temporary memorize individuals and objects, SDR-4X is equipped with long-term memory functions to memorize faces and names through more in-depth communications with people. Emotional information based on a communication experience will be memorized on a long-term memory as well. By utilizing both short and long-term memories, the SDR-4X achieves more complicated conversations and performances.
4. Speech synthesis and singing voice production
– High level of entertainment by emotionally expressive speech and voice production synchronized with full body performance
– Capability to produce singing voice with vibratos through voice synthesis by inputting music and lyric data.
To see streaming video of SDR-4X,click here.
(Available between March.19-May 31.)
5) Safe design for interaction with people
Safe design including a joint structure that does not trap hands and fingers in between joints has been adopted for safe interaction with people
6) Improved expression by adding more degrees of freedom
4 axes in the robot’s head and 1 axis in the wrist have been added in order to improve the robot’s expression. 5 individually movable fingers are also attached to each hand.
7) Improved performance of small actuators
Compared to the small actuators used in prototype “SDR-3X” announced in November 2000, a more advanced physical performance and response can be achieved by an approximately 30% improvement in the start-up torque, an approximately 15% improvement in rated torque, and an approximately 20% improvement in efficiency.
8) SDR Motion Creator
SDR Motion Creator is a software system that enables the easy production and editing of a variety of movements such as dancing, using a PC. This software also includes automatic correction functions to prevent the robot’s falling over while performing complex movements input by a creator or user. This enables the robot to easily produce complicated, personalized performances.