Font Size: a A A

Research On Motion Planning And Flocking Control For Humanoid Robot

Posted on:2016-04-10Degree:DoctorType:Dissertation
Country:ChinaCandidate:C G LiFull Text:PDF
GTID:1108330482465310Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
In the 21 st century, robots have enormous potential to greatly change people’s production and daily life. With developing of robot technology, robots will slowly enter people’s daily lives. Meanwhile, they will provide various forms of service and assistance. Humanoid robot as a hot research topic in robotics, its ultimate goal is not only to create a robot that looks like a human, but also actions, even more capable than human. In order to achieve this goal, many humanoid robot researchers and research institutions have conducted various research activities.To enhance the athletic ability of humanoid robot, meanwhile, realize intelligent formation of multi-humanoid robot and collision avoidance, the main research point of this thesis has three aspects, which include gait planning of humanoid robot, complex motion planning of humanoid robot, multi-robot formation and obstacle avoidance based on flocking control. The specific contents are as follows:1. To accomplish biped robot walking like a human, a walking pattern generation method is proposed based on natural ZMP trajectory. In the single leg support phase, based on Three Dimensional Linear Inverted Pendulum Model(3D LIPM), by setting the moving natural ZMP trajectory from heel to toe, the centroid trajectory equation is obtained. In the double leg supporting phase, a Linear Pendulum Model(LPM) is used to generate mass trajectory equation. Meanwhile, a centroid trajectory equation of multi-step planning is given in a unified coordinate. Stable walking of a biped humanoid robot, which is based on natural ZMP trajectory, is achieved in RoboCup 3D Simulation platform. Experimental results and competition results have verified the effectiveness of the proposed method.2. To achieve fast and stable shooting action, a shoot trajectory planning method based on Three Masses Inverted Pendulum Mode(TMIPM) is proposed for humanoid robots. Firstly, based on TMIPM, ZMP equations which contain swimming legs’ trajectory and torso’s trajectory are obtained. Cubic Bezier curve is used to plan swimming leg’s trajectory and ZMP trajectory, and then the torso’s trajectory of humanoid robot is calculated by ZMP equations. Secondly, in the double leg support phase, the centroid’s trajectory of a humanoid robot is calculated based on LIPM, and then achieve rapid adjustment of shooting attitude. Finally, fast shooting action of a humanoid robot is realized in the RoboCup 3D simulation platform based on the proposed algorithm. Furthermore, compare the shooting action with other teams. Experimental results show that the proposed algorithm can quickly accomplish stable shooting which just needs manually debugging alone. Moreover, the shooting time is greatly reduced, thus the competitiveness of our robot soccer team is enhanced.3. For the problem of planning robot trajectory, which is directly based on a simplified model, requires a great deal work of manual debugging. This thesis proposed a machine learning method based on Particle Swarm Optimization(PSO), which is used to optimize gait parameters and shoot action parameters. Firstly, study and analysis of gait and shooting action are made based on 3D LIPM, and then important optimized parameters are extracted. During the optimization of walking pattern, different evaluation functions are set which is dependent on gait pattern of the game as requested, furthermore, this thesis has optimized fast mode, without fall mode and high stability mode. During the optimization of shooting action, swimming leg is optimized by Bezier curve’s trajectory. The RoboCup 3D simulation platform is implemented in gait optimization and shooting optimization. In the training, the objective function is optimized layer by layer based on the tiered learning idea. Experimental results show that the tiered optimization effect of gait and shoot is better than result which is directly conducted the last optimization step.4. Exchange information is needed to be quantified in the communication network of multi-agent system. The problem of formation and obstacle avoidance is a focus research topic in robot soccer team. Therefore, this thesis is based on quantified information to do research on issues about flocking control and obstacle avoidance of a second-order multi-agent system. Assuming, a consistent quantization is used in the multi-agent system to quantify speed and position information, and a virtual leader moves along a fixed direction with uniform velocity. An input of multi-agent flocking control is designed based on quantified information in this thesis. Meanwhile, Lyapunov stability criteria and invariance principle of non-smooth system are used respectively to prove the stability in case of flock formation and in case of obstacle avoidance. Finally, Matlab simulation software is used to simulate flock formation and obstacle avoidance of the multi-agent system on a two-dimensional plane. Moreover, simulation results verify the correctness of the theoretical analysis. Based on the above algorithms, multiple humanoid robots flocking formation and obstacle avoidance behavior are realized in the RoboCup 3D simulation platform.5. The CIT3 D RoboCup simulation experimental platform has been established in this thesis. A data programming model of NAO robot has been established by using genealogy relationship. Meanwhile, a robot soccer team CIT3 D used in RoboCup 3D simulation and its debugging tools also have been established, which can be used for accurate localization. Therefore, an environment based on Particle Swarm Optimization algorithm for machine learning and training also has been established. The CIT3 D participated in several robot competitions is based on the above proposed algorithms, and has achieved good results including world silver.
Keywords/Search Tags:Humanoid Robot, Gait planning, Motion Planning, Flocking Control, Robot World Cup
PDF Full Text Request
Related items