非同轴两轮机器人自平衡与转向闭环控制

Self-balancing and Steering Closed-loop Control of Non-coaxial Two Wheeled Robot

摘要:

针对车速、车身侧倾角和前轮转角变化较大工况下的非同轴两轮机器人在基于前轮转角的自平衡控制中，因动力学模型准确性对自平衡控制带来的影响，设计了基于RBF神经网络模糊滑模控制的自平衡控制器，利用RBF神经网络的逼近特性，对动力学模型中非线性时变的不确定部分进行自适应逼近，从而提高动力学模型的准确性，并借助模糊规则削弱滑模控制中产生的系统抖振；以及因前轮转角用于自平衡控制中难以实现转向闭环控制，建立了基于纯跟踪法的轨迹跟踪控制器，并设计利用车身平衡时车身侧倾角与前轮转角的耦合关系，将转向闭环控制中的目标前轮转角替换为目标车身侧倾角，从而将自平衡控制器与轨迹跟踪控制器相结合，在保证车身平衡行驶的前提下，实现带有轨迹跟踪的转向闭环控制。实验结果表明，凭借动力学模型的较高准确性，RBF神经网络模糊滑模自平衡控制器具有鲁棒性好、超调量低和响应迅速的优点，并且利用车身平衡后车身侧倾角与前轮转角耦合关系，实现转向闭环控制是可行的，具有良好的轨迹跟踪效果。

Abstract:

Aiming at the influence of the accuracy of the dynamic model on the self-balancing control of the non-coaxial two-wheeled robot under the conditions of large changes in vehicle speed, body roll angle and front wheel rotation angle in the self-balancing control based on the front wheel rotation angle, the design a self-balancing controller based on fuzzy sliding mode control of radial basis function(RBF) neural network is proposed. Using the approximation characteristics of RBF neural network, the nonlinear time-varying uncertain part of the dynamic model is adaptively approximated, thereby improving the accuracy of the dynamic model. And the system chattering generated in sliding mode control is weakened with the help of fuzzy rules. And because the front wheel angle is used for self-balancing control, it is difficult to achieve steering closed-loop control, a trajectory tracking controller based on pure pursuit method is established, and the coupling relationship between the body roll angle and the front wheel angle when the body is balanced is designed to control the steering closed-loop control. The target front wheel turning angle in the system is replaced with the target body roll angle, so that the self-balancing controller and the trajectory tracking controller are combined to realize the closed-loop steering control with trajectory tracking on the premise of ensuring the balanced driving of the body. The experimental results show that with the high accuracy of the dynamic model, the RBF neural network fuzzy sliding mode self-balancing controller has the advantages of good robustness, low overshoot and fast response, and uses the body to balance the rear body roll angle and the front. It is feasible to realize the steering closed-loop control based on the coupling relationship of the wheel rotation angle, and has a good trajectory tracking effect.