**4. Conclusion**

10 Will-be-set-by-IN-TECH

the adaptation to using 100mm tool, in designed 3 DOF, 8 light blue circles are not on the designed trajectory and the generated trajectory(light blue line) is not matched the desired one(green line). It is to be sure that the generated trajectory represents nearer by the designed one than the changed one, satisfactory trajectory could not be obtained. Joint locking causes degrade of joint DOF, and the reachable area of end-effector has deteriorated. Even in this case, the generated trajectory is matched the optimum one acquired from exploration at all

So the robot generates a new motion by exploring in 4 DOF adding redundant waist yaw joint *θ*0. The new trajectory(blue line) in expanded 4 DOF is very nearer by the designed one than that of in 3 DOF. Because the generated trajectory is well matched the optimum one which is explored at all the 100 joint angle vectors, it is considered that some of the target positions could not be existed on reachable area even if the joint DOF was expanded to 4 DOF. It is expected that adding the other redundant joints will make the generated trajectory nearer and nearer. Average of position error is about 216mm on failure. The difference between the generated trajectory in 3 DOF and the designed one is about 30mm. The difference 7.4mm is observed by proposed method, which is accomplished about 97% similar trajectory with the

The observed acrual movements of the robot with the designed motion and the generated motions in each joint DOF are shown in Fig.8. The horizontal axis indicates the time of motion *t*. In 3 DOF, because the robot cannot bend the elbow, the robot cannot express the target

<sup>1</sup> 25 50 75 100 t

(a) The desigend motion

<sup>1</sup> 25 50 75 100 t

(b) The generated motion in 3 DOF with joint locking

<sup>1</sup> 25 50 75 100 t

(c) The generated motion in 4 DOF with joint locking

Fig. 8. The obsreved robot motions and trajectories

the 100 joint angle vectors, too.

desired one.

As a case study of the multi DOF humanoid robots, we introduced the study to execute autonomous motion adaptation against robot mechanical structure changes. To apply for humanoids which have multi DOF and sophisticated complicated structure, we proposed an autonomous motion adaptation method without model identification. We showed that the adaptation method could execute efficiency by exploring the corresponding joint angle vectors at a small number of typical points and estimating all the proper joint angles based on the result in exploration. Using proposed method, the generated trajectories are well matched the desired one on unobservable changed structure. Even if the generated motions wouldn't have achieved the designed trajectories, it is expected that the new motion achieving the desired trajectories would be generated by exploring in expanded DOF adding redundant joints.

Now, we showed experiments only for the trajectory "draw a circle". In general, humanoid robots should achieve various trajectories to provide various services. Even if the robot has several designed motions to achieve various trajectories, it is expected using VQ that the effective typical points for modifying are extracted and the designed motions can adapt to changed structure efficiently. It is needed to evaluate the effectiveness of proposed method applying for various motions. Now, we deal with only the motions based on trajectory control. This method is not limited control system. The robot chooses the evaluation function based on control system and explores the corresponding joint angle vectors using the evaluation function at typical points, the robot will generate the adapted motions for another control system. It calls for further researches and experiments. We showed the possibility to adapt motions in expanded DOF, but in this paper, adding redundant joints ware determined by experts. It is necessary to consider the joint selection method to expand joint DOF. This method realized to reduce the number of exploration points, but exploration costs increase exponentially with increasing robot joint DOF. There are some humanoid having over 30 DOF, it is difficult to explore in real time. New adaptation method which is not independent on increasing of joint DOF is needed. It calls for further researches and experiments, too.

In our method, stability problems are not resolved. On typical points, by recovering not only position sensor values but also the sensor values observing the stability such as acceleration, gyro, force and so on, the designed trajectories can be recovered in stable condition. In addition, this problems will be resolved in combination with the other methods such as Maufroy(Maufroy et al., 2010), Otoda(Otoda et al., 2009). Further studies are required.

#### **5. References**


**3** 

Riadh Zaier

*Sultanate of Oman* 

**Design of Oscillatory Neural Network for** 

**Locomotion Control of Humanoid Robots** 

*Department of Mechanical and Industrial Engineering, Sultan Qaboos University* 

Standing and walking are very important activities for daily living, so that their absence or any abnormality in their performance causes difficulties in doing regular task independently. Analysis of human motion has traditionally been accomplished by subjectively through visual observations. By combining advanced measurement technology and biomechanical modeling, the human gait is today objectively quantified in what is known as Gait analysis. Gait analysis research and development is an ongoing activity. New models and methods continue to evolve. Recently, humanoid robotics becomes widely developing world-wide technology and currently represents one of the main tools not only to investigate and study human gaits but also to acquire knowledge on how to assist paraplegic walking of patient (Acosta-M´arquez and Bradley, 2000). Towards a better control of humanoid locomotion, much work can be found in the literature that has been focused on the dynamics of the robot using the Zero Moment Point (ZMP) approach (Vukobratovic and Borovac, 2004). More recently, biologically inspired control strategies such as Central Pattern Generators (CPG) have been proposed to generate autonomously adaptable rhythmic movement (Grillner, 1975, 1985; Taga, 1995; Taga *et. al*, 1991). Despite the extensive research focus in this area, suitable autonomous control system that can adapt and interact safely with the surrounding environment while delivering

In this chapter, we deal with the design of oscillatory neural network for bipedal motion pattern generator and locomotion controller. The learning part of the system will be built based on the combination of simplified models of the system with an extensive and efficient use of sensory feedback (sensor fusion) as the main engine to stabilize and adapt the system against parameters changes. All motions including reflexes will be generated by a neural network (NN) that represents the lower layer of the system. Indeed, we believe that the NN would be the most appropriate code when dealing, to a certain limit, with the system behavior, which can be described by a set of ordinary differential equations (ODEs) (Zaier and Nagashima, 2002, 2004). The neural network will be augmented by neural controllers with sensory connections to maintain the stability of the system. Hence, the proposed learning method is expected to be much faster than the conventional ones. To validate the

The structure of the chapter is as follows: the first section will present an introduction on the conventional CPG based locomotion control as well as the Van der Pol Based Oscillator;

theoretical results, we used the humanoid robot "HOAP-3" of Fujitsu.

**1. Introduction** 

high robustness are yet to be discovered.

