Model Predictive Path Integral (MPPI) is a recognized sampling-based approach for finite horizon optimal control problems. However, the efficacy and computational efficiency of prevailing MPPI methods are heavily reliant on the quality of rollouts. This is problematic because it is hard to sample a low-cost trajectory using random control sequences, thereby leading to inferior performance and computational efficiency, especially under constrained resources. To address this issue, we propose a data-efficient MPPI method called reinforcement learning-driven MPPI (RL-driven MPPI), which significantly reduces the dependency on the quantity and quality of samples. RL-driven MPPI employs an offline-online policy learning scheme, where the offline policy learned by RL serves as the initial solution and the initial rollout generator of MPPI, effectively combining the strengths of both RL and MPPI. The rollouts generated by RL typically correspond to a lower cost-to-go compared to random sampling, which significantly boosts the sample efficiency and convergence speed of MPPI. Moreover, the value function learned by RL offers an accurate estimation for infinite-horizon cost-to-go, enabling it to serve as a terminal term for the cost criteria of MPPI. This approach empowers MPPI to approximate an infinite-horizon cost with a shorter prediction horizon, thus enhancing real-time performance at each time step. An unmanned aerial vehicle control task is conducted to evaluate the proposed method. Results indicate that the proposed RL-driven MPPI method exhibits superior control performance and sample efficiency.

This work provides a generalization of the three-dimensional velocity obstacle (VO) collision avoidance strategy for nonlinear second-order underactuated systems in three-dimensional dynamic uncertain environments. A hierarchical architecture is exploited to deal with conflicting multiple subtasks, which are defined as several rotations and are parameterized by quaternions. An improved VO method considering the kinodynamic constraints of a class of fixed-wing unmanned aerial vehicles (UAV) is proposed to implement the motion planning. The position error and velocity error can be mapped onto one desired axis so that, only relying on an engine, UAVs can achieve the goal of point tracking without collision. Additionally, the performance of the closed-loop system is demonstrated through a series of simulations performed in a three-dimensional manner.

Data-driven learning optimization, which considers optimization as a means to perform end-to-end learning, is an emerging methodology used to solve large-scale learning and continuous control tasks. These methods provide mathematically tractable solutions with inherent interpretability, but their training can be inefficient due to the need to differentiate through the optimization problem and to solve gradient information in every training iteration. To address this issue, we propose a novel learning algorithm called Efficient Differentiable Model Predictive Control (EDMPC), which significantly improves computational efficiency by integrating an optimization-driven inference step within the backpropagation process. More concretely, we derive an analytical gradient of the trajectory from the optimal problem relative to the tunable parameter, enabling significant acceleration of the end-to-end learning procedure. This gradient can be solved by a standard optimizer in a recursive form, which then provides the direction for parameter updates. An unmanned aerial vehicle control task was conducted to demonstrate the effectiveness of EDMPC in controlling complex dynamic systems. The results indicate that the proposed EDMPC method exhibits superior efficiency in the learning process and accuracy in the control process.

It is well-known that the successful implementation of current sliding-mode guidance laws relies on two key points: one is the prior knowledge of the upper bound of the target’s maneuver, and the other is the accurate measurement of motion information. The truth maneuver is usually difficult to acquire and can only be roughly estimated. Measurements are often affected by noise and physical limitations of the onboard sensor. This paper presents an improved second-order sliding-mode guidance law that can handle these two problems simultaneously while addressing the chattering phenomenon, which inherently exists in the sliding mode-based controller. We achieve this by using the estimation from an extended state observer (ESO) and tracking differentiator (TD). In light of the employed ESO, which is utilized to give an estimation of the external disturbance, prior knowledge of the target’s maneuver is not required. Most interestingly, we show that the motion information can be estimated efficiently from the ESO in real time. A TD filter can be further embedded to remove unwanted noise from the detected signal. A series of simulations has been conducted, clearly demonstrating that the estimation accuracy of the ESO is sufficient for the guidance law to implement accurate interception. The TD filter not only removes noise but also avoids the phase-loss problems associated with conventional filters. Moreover, the chattering phenomenon is completely eliminated in the control channel.