A multi-microprocessor system to calculate the joint angles of a 6 axis manipulator from the handposition and orientation is discussed. The system is composed of one personal computer (6809) and seven Z 80 CPUs accompanied with APUs. It has common memories whose access time is very short and a circuit to synchronize the CPUs against the variation of task processing time.The system performs the parallel processing of MIMD (Multiple Instruction stream and Multiple Data stream) type.
A method to improve the joint angle calculation so as to be easily performed by parallel processing is proposed. A heuristic scheduling method for task processing which is easily performed even by the personal computer is also proposed.
It is verified that the calculation of the joint angles is performed with the computing time of 3.3 ms by the proposed system and methods while it takes 20ms by a CPU and a APU in an ordinary method.

View full abstract

The prototype four-legged vehicle for a locomotion system of an intelligent robot was described. For the legs, the horizontally articulated type manipulators with three degrees of freedom were disposed at the corners of a square frame, and whole mechanism was constructed of carbon fiber reinforced plastic (CFRP) for weight reducing purpose. The microprocessor-based hierarchical structure was used for the control system and the software was stated in C language. Forward and backward walking was performed using the walking sequence“crawl gait”which is stored as a table memory. The walking speed of 8 cm/s could be realised.

View full abstract

In the present manufacturing systems, the path control of robots is performed by teaching/playback method. However, this method is not feasible once the content of the work of the robot is altered, since we must repeat trial and error so that the motion satisfies specified performance.
For this problem, there are many approaches to establish the off-line teaching method. We here propose a path control of robots based on linearlized constraint control which has been proposed by authors, previously. Using simulation, we confirm that this method has good performances and consider that it should be directly applicable to the present manufacturing robot systems.

View full abstract

Walking vehicle has potential capability to be developed into off-road vehicle with high mobility and adaptivity by using the coordination control of its multi-degrees of freedom. There have been several discussions on the gait control of the walking vehicle. However the overall discussion of the gait control system and discussion on the subsystem in relation to all has not been executed so far.
The paper thus try to discuss this problem by setting the quadruped walking vehicle, which the authors have been developing. As the direct objective of the study. First of all the premises for discussion are established. The hierarchical structure of the total control system consisting of three level. i.e. level A, B, C, are then clarified. The control algorithm of each level especially on level B, C are elaborated. As to level B, the gait control in xy coordinates, that in z coordinates, and trajectory control of the legs are discussed. As to level C the control algorithm of the reflex motion regulation are discussed. Finally these discussions are verified by the walking experiments of the constructed walking vehicle model TITAN III. The joystick control of omnidirectional motion., or adaptive locomotion over irregular steps are successfully demonstrated.

View full abstract

The purpose of this study is to accomplish“Dynamic Walking”by a biped walking robot WL-10-RD on a disturbance-free flat floor. Generally, one walking cycle is divided into two phases, a“Single Support Phase”and a“Change Over Phase”.We consider that“Dynamic Walking”is defined as walking to which“Systen Dynamics”is applied on“Both Phases”.
We propose a dynamic walking control method as follows: walking in single support phase is rea-lized by“Program Control”using a“Preset Walking Pattern”and that in change over phase is achieved by“Sequence control”changing“Preset Torque and Mechanical Impedance Values”according to the conditions of this phase.
Practically, for the single support phase, the“Preset Walking Pattern”is designed so that the trajectory of“Zero Moment Point”on the floor plane, which is computed using“3-Dimensional Dynamic Simulation”, will be exirt inside the support sole's stable area. For the change over phase, which is divided into 4 phases, adequate values of torque and mechanical impedance on both ankles for each sub-phase are defined so that the“Center of Gravity”will chage support legs and will suppress an impact caused by floor contact.
As a result of experiments, complete dynamic walking was realized by a walking robot WL-10RD. The walking time is about 1.5 second with 40cm step.

View full abstract

Advanced robots will have to not only have‘hard’functions but also have‘soft’functions. Therefore, the purpose of this study is to realize‘soft’functions of robots such as dexterity, speediness and intelligence by the development of an anthropomorphic intelligent robot playing keyboards instrument. This paper describes the development of keyboard playing robot WABOT-2 (WAseda roBOT-2) with a focus on the mechanisms of 4 limbs and their control system.
This robot has two arms with 5 fingers each, and two legs for pedaling, totaling 50 degrees of freedom (DOF) . There is no precedent for the development of the robot system with such many DOF, and synthesis of computer system capable of efficient control of those multi-degrees of freedom emerges as a big problem. And, it is also the most important for this autonomic limb control system, to own the capability to produce the trajectories of 4 limbs from the score information input.
To deal with above problems, the computer system of the WABOT-2's limb control system is hierarchically structured by three levels wi th 53 micro-processors. The top level processes the musical score information, and the processing flow can be divided into two steps. First, by using a problem solving method of artificial intelligence, fingering and wrist positions are determined according to the score data sent from the vision system. Second, the trajectories of fingers, arms and legs are planned. These processes are based on three types of knowledge; object, work and the specifications of the robot. The middle level sends the reference position to each degree and realizes coordination control according to the trajectory plan by monitoring the condition of each degree when required. At the bottom level, a single chip micro-computer is allotted to each DOF, providing software servo-mechanism. This computer system has made it possible to control 50 DOF efficiently.
As a result, WABOT-2 system has realized fluent play of tunes of the intermediate electronic-organ textbook on a keyboard instrument only by giving information on musical score.

View full abstract

The Waseda University project team built up an intelligent robot WABOT-2 which can play an electric piano, using ten fingers and feet, while reading printed music. It can hold a conversation with a man using an artificial voice.
This paper reports on its vision system, which can recognize not only a commercially available printed score but also instant lettering score in approximately 15 seconds, using an extraction algorithm based on the music score structure with special hardware and knowledge base music understanding software.
Several automated recognition system for printed music have been reported However, none of them are completely adequate for use as music playing robot vision, where real time recognition is required under an insufficient data environment where a video camera is placed in the robot's head and music sheet is often set on the music stand in its curved or deformed style.
In order to directly convey the vision system output to the robot hands, contradictions and inharmonic codes in the pattern discrimination results are checked, using knowledge base musical syntax. Moreover, pattern interpolations for drop-out are performed by the relaxation method at the discrimination level. Music information interpolations for missing discrimination are also achieved, based on harmonic codes and music grammar at the following understanding level.
The resulting music robot vision performance is sufficient, wherein one sheet of commercially available printed of nursery song music for an electric piano with three parts can be recognized in approximately 15 seconds, with 100% accuracy.

View full abstract

If ROBOT has the ability to communicate with human by flexible conversation, it becomes more human-like-being and the field of application will be considerably expanded.
In this paper, a speech input/output system to achieve these purpose, which is implemented as a conversational sub-system of WABOT-2 who is a key-board player, is described.
The system is composed of two parts: One is the conversational speech understanding part to recognize the continuously uttered sentences appropriately according to the situation, the other is the speech synthesis part to prompt the speech input to human naturally and to realize the smooth and natural conversation.
To realize the flexibility of the task, the phoneme-recognition-based-system and the synthesis-by-rule using CV-syllable are adopted in understanding part and synthesis part, respectively. In these cases, the change of the sounds according to the surrounding phonemic context, so called coarticu-lation, becomes main problem in understanding part, and the deterioration of the quality of the synthesized speech becomes that in synthesis part. In this study, the effective use of the articulatory feature is proposed to compensate coarticulation. As for the problem of synthesis part, the sensitivity of the k-parameter to the formant frequency is considered.
One more important problem to construct such a system is how to consider the topic of the conversation. In this system, a network model having path-weighting-facility and a weight-control-scheme based on the production rule are proposed for the sentence recognition. These method enable to consentrate the subset of the network representing the acceptable sentences according to the situation, and to understand uniquely even if the sentence is ambiguous one having multiple meanings.
These algorithms are implemented on the hardware with 15 microprocessors and are able to run in real time.
With these products, the conversational system with high flexibility is realized.

View full abstract

This paper gives an overview of the keyboard playing robot WABOT-2 and presents in some detail its supervisory system and the singing voice tracking subsystem. The WABOT-2 is an anthropomorphic robot which is made up of four subsystems; a supervisory system, located functionally atop the four subsystems, has the function of monitoring and controlling the operation of the respective subsystems. In an attempt to realize successful control of the subsystems, a“mimic”WABOT-2 capable of simulating the operation of each subsystem was constructed prior to the completion of the actual subsystems, and problems such as communication commands, codes of musical notes, the diagnostics for a subsystem malfunction were investigated, and the procedure of control was represe-nted by rules for the purpose of extension in the future.The results of this simulation are reflected in the design of the supervisory and communication systems of WABOT-2.
The singing voice tracking subsystem detects deviations of singing voice in the accompaniment of the WABOT-2. Codes of musical notes are generated on-line and compared with nominal score. Any deviation in excess of a prescribed threshold is detected; the result is reflected in the replay as a change of key. Also from generated musical notes this system can also guess a song sung by a person.

View full abstract