【作者】 高志伟；

【导师】 代学武；

【作者基本信息】 东北大学 ， 控制理论与控制工程， 2018， 硕士

【摘要】 随着移动机器人技术的进一步发展以及自动驾驶、智能快递等新兴领域的崛起,移动机器人面对的工作任务越发复杂,在有限的车载电池下机器人所能实现的工作时间和运行范围不足以完成这些任务。如何减少系统能耗,尽可能高效地完成指定任务,是提高客户满意度、减少运行成本的一种重要方法,也顺应了全球节能减排的大环境。其中通过路径规划的方法使机器人以最小的能耗完成移动任务到达指定目的地,是实现移动范围扩大和延长工作周期的核心技术,具有良好的研究前景。但是对于不同类型、不同功能的移动机器人,现有的能耗优化轨迹规划方法各不相同,有的方法忽略了传感器、控制电路等造成的非运动能耗,没有对运行时间进行优化;有的方法只单独的以距离优化或轨迹光滑优化为标准,没有考虑与运行距离有关的摩擦能耗和与轨迹光滑度有关的状态改变耗能两者间互为矛盾的关系;且大多数方法将路径规划与轨迹跟踪控制单独考虑,忽略了由控制器的动态响应造成的实际轨迹与期望轨迹能耗并不完全相同,使得建立的能耗模型并不准确。以及现有方法大多是在所建能耗模型的基础上,通过对轨迹一一搜索或迭代优化得到最优轨迹的,所需计算量巨大、求解困难。针对上述原因造成的能耗模型不准确、优化求解困难等共性问题,本文以两轮自平衡机器人为研究载体,考虑到控制器的动态特性对能耗的影响,将轨迹规划与由轨迹跟踪控制器和机器人动力学方程组成的运动控制模型相结合,提出了一种基于频域分析的、可准确预测机器人耗能的能耗模型。并面向平坦的、信息完全已知的运行环境,提出了一种可快速求解的能耗最优路径规划算法,以此来提高移动机器人系统的节能性。为了降低地面摩擦系数的变化对本算法准确性的影响,采用在线动态辨识策略,动态更新能耗模型和控制器的参数,提高了算法对环境的鲁棒性,保证了所得轨迹的能耗最优性。现对文中的主要成果总结如下:(1)对于轨迹跟踪控制器,将经典的LQR和PID算法相结合,采用了一种LQR-PID最优平衡和轨迹跟踪算法。该算法综合了两者优点,具备良好的动态特性和静态特性,且具有较少的控制能耗。并通过仿真实验证明了该控制算法可以使机器人保持平衡的同时,良好地跟踪设定的系统状态,具有较强的通用性和抗干扰能力。(2)对于能耗模型,通过对经典能耗最优轨迹规划算法的优缺点分析,得出只在直线轨迹或完全光滑轨迹的框架下考虑能耗优化是不够的,所以将两者结合,提出了一种新的期望轨迹表达式。考虑到由于控制器的动态特性,机器人实际轨迹与由轨迹规划得到的期望轨迹能耗并不完全相同,依据帕塞瓦尔定理,提出了一种基于频域分析的机器人能耗模型。该模型充分考虑了控制器的动态响应和传感器等非机械部件的能耗,能够准确反映期望轨迹与机器人能耗间的关系。(3)对于能耗最优轨迹规划,分别对全局路径规划算法和局部轨迹规划算法进行了能耗优化设计。对于全局路径规划,用与能耗相关的标准替换A*算法中的距离标准,并考虑了与障碍物的距离因素;对于局部轨迹规划,基于圆弧轨迹各状态间的关系,通过一定的数学变换将能耗模型降维为只含轨迹对应圆心角和运行时间两个变量的函数,再通过简单地数学求导可直接得到最优轨迹,不需要对符合要求的轨迹一一搜索或迭代优化,大大加快了求解速度。考虑地面摩擦系数的变化对能耗模型和控制器的影响,提出了一种在线动态辨识的策略,利用动态加权最小二乘法实现能耗模型和控制器参数的动态更新,提高了算法对环境的鲁棒性。(4)搭建了基于MATLAB/Simulink的仿真平台,开展了对上述方法的实验研究。实验结果表明,本文所提能耗模型准确率达到99%以上,能准确地反映系统的运行能耗,相比于常用的能耗最优三次贝塞尔曲线法和最短距离法,本文所提的能耗最优轨迹规划算法的节能率分别提高7.8%和4.5%,节能效果显著提高。并结合LEGOEV3两轮自平衡机器人和无线定位系统搭建了实物平台,为之后的实物实验提供了条件。更多还原

【Abstract】 With the further development of mobile robot technology and the rise of new fields such as autonomous driving and intelligent express delivery,mobile robots are faced with more and more complex tasks.Under the limited on-board battery,the working time and operating range of the robot are not enough to complete these tasks.How to optimize the energy consumption of the system and complete the assigned tasks as efficiently as possible is an important method to improve customer satisfaction and reduce operating costs.It also conforms to the global environment of energy conservation and emission reduction.Among them,the path planning method enables the robot to complete the moving task with minimum energy consumption to reach the designated destination,which is the core technology to realize the expansion and extension of the moving range and work cycle,and has a good research prospect.However,for mobile robots with different types and functions,the existing energy consumption optimization trajectory planning methods are different.Some methods ignore the non-moving energy consumption caused by sensors and control circuits and do not optimize the running time.Some methods only take distance optimization or trajectory smooth optimization as the standard,without considering the contradictory relationship between friction energy consumption related to running distance and state change energy consumption related to trajectory smoothness.Moreover,most methods consider path planning and trajectory tracking control separately,ignoring that the actual trajectory and the expected trajectory energy consumption caused by the dynamic response of the controller are not exactly the same,making the energy consumption model established inaccurate.In addition,most of the existing methods are based on the established energy consumption model to obtain the optimal trajectory through one-by-one search or iterative optimization of the trajectory,which requires a huge amount of calculation and is difficult to solve.Aiming at the common problems such as inaccurate energy consumption model and difficult optimization solution caused by the above reasons,this paper takes two-wheeled self-balancing robot as the research carrier.Considering the influence of the dynamic characteristics of the controller on energy consumption,the trajectory planning is combined with the motion control model composed of the trajectory tracking controller and the dynamic equation of the robot and an energy consumption model based on frequency domain analysis which can accurately predict robot energy consumption is proposed.And an energy consumption optimal path planning algorithm is proposed for flat and fully known operating environment,so as to improve the energy saving of mobile robot system.In order to reduce the influence of the variation of ground friction coefficient on the accuracy of the algorithm,the on-line dynamic identification strategy was adopted to dynamically update the energy consumption model and the parameters of the controller,so as to improve the robustness of the algorithm to the environment and ensure the optimal energy consumption of the obtained trajectory.The main achievements are summarized as follow:(1)For the trajectory tracking controller,a LQR-PID optimal balance and trajectory tracking algorithm is proposed by combining the classical LQR and PLD algorithm.This algorithm combines the advantages of both methods,which has good dynamic and static characteristics and has less control energy consumption.The simulation results show the control algorithm can keep the robot in balance and track the set system states well and it has strong universality and anti-interference ability.(2)For the energy consumption model,by analyzing the advantages and disadvantages of the classical energy consumption optimal trajectory planning algorithm,it is concluded that it is not enough to consider energy consumption optimization only under the framework of linear trajectory or completely smooth trajectory.Therefore,a new expected trajectory expression is proposed by combining the two.Considering the dynamic characteristics of the controller,the energy consumption of the actual trajectory is not exactly the same as the expected trajectory obtained by the trajectory planning,so an energy consumption model based on frequency-domain analysis is proposed based on parseval’s theorem.The model takes full account of the dynamic response of the controller and the energy consumption of non-mechanical components such as sensors,and can accurately reflect the relationship between the desired trajectory and the energy consumption of the robot.(3)For the optimal energy consumption trajectory planning,the global path planning algorithm and the local trajectory planning algorithm are designed respectively.For the global path planning,the distance criteria in A*algorithm is replaced with the criteria related to energy consumption,and the distance factor for obstacle is considered.For the local trajectory planning,based on the relationship between the states of the arc trajectory,the dimension of the energy consumption model is reduced to a function containing only the two variables of the trajectory corresponding to the central Angle and the running time through certain mathematical transformation.Then,the optimal trajectory can be obtained directly by simple mathematical derivation,without searching or iterative optimization for the trajectories that meet the requirements one by one,which greatly accelerates the solving speed.Considering the influence of ground friction coefficient on energy consumption model and controller,an online dynamic identification strategy is proposed.The energy consumption model and controller parameters are dynamically updated by using the dynamic weighted least square method,which improves the robustness of the algorithm to the environment(4)The simulation platform based on MATLAB/Simulink was established and the experimental study of the above method was carried out.Experimental results show that the accuracy of the energy consumption model proposed in this paper is more than 99%,which can accurately reflect the energy consumption of the system.Compared with the commonly used energy consumption optimal cubic Bessel curve method and the shortest distance method,the energy saving rate of the energy consumption optimal trajectory planning algorithm proposed in this paper is increased by 7.8%and 4.5%respectively,and the energy saving effect is significantly improved.In addition,the LEGO EV3 two-wheeled self-balancing robot and wireless positioning system are combined to build a physical platform,which provides conditions for the subsequent physical experiments.更多还原