Chenxu Ke

Chenxu Ke, Congling Tian, Kaichen Xu et al.

Reinforcement learning-based controller design methods often require substantial data in the initial training phase. Moreover, the training process tends to exhibit strong randomness and slow convergence. It often requires considerable time or high computational resources. Another class of learning-based method incorporates Lyapunov stability theory to obtain a control policy with stability guarantees. However, these methods generally require an initially stable neural network control policy at the beginning of training. Evidently, a stable neural network controller can not only serve as an initial policy for reinforcement learning, allowing the training to focus on improving controller performance, but also act as an initial state for learning-based Lyapunov control methods. Although stable controllers can be designed using traditional control theory, designers still need to have a great deal of control design knowledge to address increasingly complicated control problems. The proposed neural network rapid initialization method in this paper achieves the initial training of the neural network control policy by constructing datasets that conform to the stability conditions based on the system model. Furthermore, using the image-based visual servoing control for multicopter interception as a case study, simulations and experiments were conducted to validate the effectiveness and practical performance of the proposed method. In the experiment, the trained control policy attains a final interception velocity of 15 m/s.

Xunhua Dai, Chenxu Ke, Quan Quan et al.

Control systems on unmanned vehicles are safety-critical systems whose requirements on reliability and safety are ever-increasing. Currently, testing a complex autonomous control system is an expensive and time-consuming process, which requires massive repeated experimental testing during the whole development stage. This paper presents a unified simulation and test platform for vehicle autonomous control systems aiming to significantly improve the development speed and safety level of unmanned vehicles. First, a unified modular modeling framework compatible with different types of vehicles is proposed with methods to ensure modeling credibility. Then, the simulation software system is developed by the model-based design framework, whose modular programming methods and automatic code generation functions ensure the efficiency, credibility, and standardization of the system development process. Finally, an FPGA-based real-time hardware-in-the-loop simulation platform is proposed to ensure the comprehensiveness and credibility of the simulation and test results. In the end, the proposed platform is applied to a multicopter control system. By comparing with experimental results, the accuracy and credibility of the simulation testing results are verified by using the simulation credibility assessment method proposed in our previous work. To verify the practicability of the proposed platform, several successful applications are presented for the multicopter rapid prototyping, estimation algorithm verification, autonomous flight testing, and automatic safety testing with automatic fault injection and result evaluation of unmanned vehicles.