Research On For A Two-wheeled Self-balancing Vehicle Trajectory Tracking Key Technology

Wheeled inverted pendulum (WIP) vehicles have became widely available and continuously received much attentions in intelligent transportation, space exploration and domestic service, owing to its small size, simple structure, high mobility, and low energy characteristics. At the same time, as the advanced form of inverted pendulum system, the motion control of the WIP vehicle has a significant research value for the robotics and related control fields. When applied in unmanned vehicle or space exploration, the WIP vehicle needs to achieve trajectory tracking. On this background, this paper launches the research work considering various circumstances.Firstly, based on the mathematic model and the zero dynamic concept in control theory, the unstable characteristic of the WIP vehicle caused by the exceptional mechanical structure is analyzed in details. Then the feedback linearization technology and Lyapunov stability theory are used to illustrate the root of unstable dynamics. The Olfati transformations decouple the underactuated coupling state variables and transform the dynamic model into strict feedforward cascade system, which possesses the stable zero dynamic. Based on this, an overall sliding mode control which considers all the underactuated states is designed to deal with the underactuated longitudinal motion. The terminal sliding mode control makes the rotational motion be rapidly convergent in finite time. Then by the aid of model predictive control (MPC) which generates the desired smooth trajectory, the WIP vehicle can achieve the arbitrarily trajectory tracking in the earth frame.Secondly, taking the various disturbances and physical constraints, such as actuator saturation and vehicle body tilt angle, into account, a hierarchical control strategy based on MPC is presented to handle the trajectory tracking. An outer-loop saturated kinematic controller with global stabilization is introduced to generate a virtual velocity, which guarantees the tracking error converges to a small neighborhood of the origin and deals with the nonholonomic constraint problem. As the inner-loop dynamic controller, the model predictive control is based on the feedback linearization model and employed to handle a comprehensive issue associated with underactuated problem, state limitations and various disturbances, where the nonlinear optimization problem is transformed as a standard quadratic programming (QP) problem. MPC ultimately asymptotically approaches to the real-time generated virtual velocity, and ensures the vehicle body sustain around vertical position in the in the meantime.Finally, considering the inaccessible accurate model and immeasurable velocity signal, a controller based on the extended state observer is proposed in the presence of external disturbances and physical constraints. An extended state observer is established to estimate the velocity variables and disturbances including the external disturbance and model uncertain dynamics, which is independent of the controller and can be directly utilized to the controller as the state feedback and feed-forward compensate. The dynamic controller integrates the adaptive sliding model control and model predictive control An adaptive sliding model control is designed for the rotational velocity tracking, where the adaptive law can reject the uncertain parameters and observation errors. The MPC is directly utilized to the nonlinear model and achieves longitudinal velocity tracking and vehicle body stability, which does not need feedback linearization and reduces dependence on accurate system dynamic model.This paper can significantly improve the WIP tracking precision and vehicle body stability in the earth frame, and also reduce the influence of both internal model uncertainty and external disturbance. In addition, the controller can be applied in the case where the sensors can not acquire vlocities informations. The various simulations demonstrate the effectiveness of the proposed control strategy.