摘要
将特征点的深度信息和像素坐标作为视觉特征,提出一种视觉伺服准最小最大模型预测控制(MPC)方法.与传统方法相比,机器人摔制信号可通过在线求斛线性矩阵不等式的凸优化问题获得,其可行斛可保证系统的闭环渐近稳定性.该方法易于处理系统约束,在满足执行器机械限制的前提下能够有效规划特征点的图像轨迹.同时,深度特征的引入对于改进摄像机的三维轨迹具有显著效果.六自由度工业机器人手眼系统的仿真结果验证了所提出算法的有效性.
Taking depth information and pixel coordinates of the feature points as image features, a quasi-min-max modelpredictive control(MPC) algorithm for image-based visual servoing is presented. Compared with the traditional method, therobot control signals can be obtained by the convex optimal problem involving linear matrix inequalities(LMIs), and theclosed-loop stability of visual servoing system is guaranteed by the feasibility of the LMIs. The proposed method is easy todeal with the system constraints. Under the premise of actuator mechanical limitations, the image trajectories of the featurepoints are effectively constrained. Furthermore, the introduction of the depth information significantly improves the three-dimensional trajectory of the camera. The simulation results on a 6 degrees-of-freedom robot manipulator with eye-in-handconfiguration show the effectiveness of the proposed algorithm.
出处
《控制与决策》
EI
CSCD
北大核心
2013年第7期1018-1022,共5页
Control and Decision
基金
围家白然科学基金项目(60804013)
关键词
视觉伺服
深度信息
约束
线性矩阵不等式
准最小最大模型预测控制
visual servoing
depth information
constraint
linear matrix inequalities(LMIs)
quasi-min-max modelpredictive control(MPC) algorithm