This paper introduces a hybrid algorithm of deep reinforcement learning (RL) and Force-based motion planning (FMP) to solve distributed motion planning problem in dense and dynamic environments. Individually, RL and FMP algorithms each have their own limitations. FMP is not able to produce time-optimal paths and existing RL solutions are not able to produce collision-free paths in dense environments. Therefore, we first tried improving the performance of recent RL approaches by introducing a new reward function that not only eliminates the requirement of a pre supervised learning (SL) step but also decreases the chance of collision in crowded environments. That improved things, but there were still a lot of failure cases. So, we developed a hybrid approach to leverage the simpler FMP approach in stuck, simple and high-risk cases, and continue using RL for normal cases in which FMP can't produce optimal path. Also, we extend GA3C-CADRL algorithm to 3D environment. Simulation results show that the proposed algorithm outperforms both deep RL and FMP algorithms and produces up to 50% more successful scenarios than deep RL and up to 75% less extra time to reach goal than FMP.

A path following controller based on an uncertainty and disturbance estimator (UDE) for a quadrotor with a cable suspended payload is proposed. The quadrotor and the payload are subject to unknown wind disturbances. The controller resembles a cascade architecture. For the outer loop, a UDE-based translational control law is proposed. The controller asymptotically stabilizes the quadrotor along a given path and estimates the lumped disturbances with a low pass filter. For the inner loop, an attitude tracking controller is used to control the direction of the lift vector so that the actual lift force can asymptotically follow the reference force generated by the translational controller. The stability of the system with the translational controller and the attitude tracking controller has been shown to be asymptotically stable using the reduction theorem. With the help of the reduction theorem, the design of the translational and the attitude control can be decoupled, providing the flexibility of implementing different attitude controllers without redoing the stability analysis. As shown in the simulation, the control law can stabilize the quadrotor on the desired path under different wind disturbances.

This paper presents a novel guidance and control design for a parcel tethered to a drone in its delivery mission. The proposed proportional navigation based guidance, integrated with custom built control, enable us to achieve autonomous payload landing. The novelty of the proposed method lies on two aspects: 1) the guidance law allows for soft landing; while 2) the path following control ensures the swing-free payload transportation that sets the solid foundation for landing. The development is verified by extensive simulations and further demonstrated by flight experiments.

The maintenance of visual features within the sensor field of view poses a significant challenge for underactuated aerial vehicles such as quadrotors, especially during aggressive maneuvers. However, the existing image-based visual servo (IBVS) control methods rely on strict target visibility assumptions or impose excessive constraints on the quadrotor's agility to meet this requirement. Furthermore, the effectiveness of the visibility constraint defined in prior works remains unverified in aggressive flight tests. To address these issues, we present a robust IBVS scheme for quadrotors to perform aggressive maneuvers while ensuring target visibility. Based on the nonlinear model-predictive control framework, we propose a novel underactuation compensation scheme to eliminate the need for a virtual camera frame, which enables us to formulate the true target visibility constraint. We then introduce a quaternion-based representation of spherical visual features to handle the nonsmooth vector field problem on the 2-sphere and derive its corresponding image kinematics. We validate our method through three challenging visual servo tasks where agile maneuvers are desired: fast landing, aggressive long-distance flight, and dynamic object tracking. Extensive simulation and experiment show that our method consistently achieves a target-visible rate of 100% in all the image frames, even under a maximum pitch of 21.04 $^{\circ }$ . The results validate the effectiveness of our visibility constraint under large robot rotations and underscore its importance in enabling robust and aggressive flights.