|
Humanoid robots are one of the most popular research topics in the robotic society since their mechanisms have high capabilities to perform dexterous and versatile tasks in human environments. With the improving computer technology in the early of 70s, humanoid robots have been extensively studied due to their several superiorities such as adaptation abilities to human environments and performing tasks that people can do and even tasks that people are unable to do. Humanoid robots have high potentials to improve the quailty of humans' lives, e.g., they can be used to take care of kids and/or old people and perform tasks in a hazardous environment, in which humans are not willing to work, such as Fukushima Nuclear Plant and Soma mine pit. The main motivations behind the humanoid researches can be stated as follows: against wheel and caterpillar, humanoid robots can locomote in a discontinous terrain efficiently; the mechanisms of humanoid robots are very suitable to perform general tasks in human environments; it can be implemented into several different fields such as military and entertainment; psychological factors in human-robot interaction and so on. Besides, humanoid researches provide deep insight into the nature of humans' motions which can be used to design new mechanisms that can help humans, e.g., human support systems are used to improve the motion skills of old and/or disabled people.
Although humanoid robots have high capabilities to perform complex tasks in human environments and there are several motivation sources, which are explained above, behind the humanoid researches, the existing ones are still far away from being a part of our daily life due to some challenging issues such as safety. Humanoid robots should be able to adapt to unknown environments to perform dexterous and versatile tasks in human environments. For instance, humanoid robots are generally walked on even terrains; however, human environments include uneven terrains as well.
Bipedal locomotion has been widely studied in humanoid researches; however, upper body control of humanoid robots is as important as lower body control to perform humans' daily life activities. For instance, holding a glass of water, showing a direction, and pushing an object require upper body motion control. Therefore, not only the lower body control but also the upper body control should be studied to improve the versatility of humanoid robots.
In this dissertation, whole body dynamic model and control of humanoid robots are studied to perform dexterous tasks in human environments. The exact dyanmic model of a humanoid robot is derived by using floating point base dynamics. A new simulator is designed by using the derived whole body dynamic model and Virtual Reality Toolbox of MATLAB. Several different control tasks, in which a humaonid robot works in an unknown open environment, are conducted by using the proposed simulator.
This dissertation includes six chapters. The organization of the dissertation is as follows:
In the first chapter, the introdcution of the dissertation is given. A literature review is provided for bipedal locomotion and humanoid researches. The main motivation of humanoid researches and the demerits of existing humanoid robots are explained in detail. The problem definition and the contribution of the dissertation – improving the dexterity and versatility of humanoid robots by analyzing the dynamic characteristics of humans' motions and designing advanced controllers- are explained clearly. The fundamental concepts of bipedal locomotion and humanoid robots are explained.
In the second chapter, advanced motion control methods, which are used in the whole body control of humanoid robots, are briefly explained. Disturbance observer is used to achieve high performance robust motion control systems. Disturbance observer estimates external disturbances as well as system uncertainties, and the robustness of a motion control system is simply achieved by feeding back the estimated disturbances. If system uncertainties are substracted from the input of a disturbance observer, then external disturbances can be estimated, i.e., disturbance observer works as a force sensor. Force sensorless force control is significantly important in humanoid applications since there may be several contact points on a humanoid robot in contact motion. However, although joint space forces/torques can be easily estimated by using a disturbance observer, contact points should be known a priori to control operational space forces. A compliance motion control system is used to treat impact force when swing foot contact to floor. The compliance control is very important for the balance of humanoid robots. If it is not used, then robot may directly tip over due to the impact forces at sole. A hybrid motion control system is also proposed to achieve contact motion control in the upper body of humanoid robot.
In chapter three, the dynamic model of a humanoid robot, which has floating point base dynamics, is derived by using two different methods. The floating point base dynamics is used so that the continous dynamic model of a humanoid robot is derived. In the first method, the dynamic model is numerically derived by using generalized Newton-Euler algorithm. Although the dynamic model is systematically derived, it does not give deep insight into the dynamic characteristics of humanoid robots. In the second method, the fundmentals of floating point base dynamics are considered and the dynamic model is analytically derived. It provides better insight, yet Coriolis and centrifugal forces cannot be derived.
In chapter four, lower body control of humanoid robots is considered by analyzing a two legged bipedal mechanism. For the sake of simplicity rotational motion about yaw axis is ignored. Static and dynamic walking simulations of the two legged bipedal mechanism are given by using the proposed simulator. It is shown that the speed of bipedal locomotion can be increased by using dynamic walking. Zero moment point (ZMP) is used to achieve the balance in dynamic locomotion. ZMP is widely used to generate stable dynamic walking pattern on even terrain. However, humanoid robots should be able to walk on uneven terrains to adapt human environment. Virtual ZMP and virtual support polygon are used so that the two legged bipedal mechanism can dynamically walk on an uneven terrain. The main disadvantage of the virtual ZMP is that the uneven terrain should be known to determine virtual support polygon. Therefore the applications of virtual ZMP is limited.
In chapter five, whole body motion control of humanoid robots is studied. The whole body motion control of a humanoid robot consists of lower and upper body control systems. The lower body control is studied in chapter 4. In this chapter, two different states of the upper body control, namely free and contact motion control, are considered. In the free motion control, robot arms do not contact any object and position controllers are designed so that they follow predetermined trajectories. By keeping arm accelerations low, upper and lower body controllers of a humanoid robot can be idependantly designed. It is shown that a humanoid robot can perform free motion tasks and keep its balance by using the propopsed controllers. In the contact motion control, the robot arms contact to unknown environment. To achieve contact motion, hybrid controllers are conducted in the upper body control. The stability of humanoid robot is significantly influenced by the stability of contact motion. If the impact force is high and/or the stability of contact motion is low, then the humanoid robot may direclty tip over. In order to improve the stability of humanoid robots, safe contact motion control should be achieved by improving the adaptibility of humanoid robots. However, designing a high performance humanoid robot that can adapt unknown open environment is not an easy task. A bilateral control system, in which humans' skills can be transferred to humanoid robots, is implemented in the control of humanoid robots so that the adaptibility is improved. Two different bilateral control systems are implemented in the control of humanoid robots. Several different dexterous and versatile applications can be performed by using the proposed control methods.
The dissertation ends with conclusion and discussion given in the sixth chapter. |
