Prediction Error-Based Action Policy Learning for Quadcopter Flight Control

Quadcopters are finding their place in everything from transportation, delivery, hospitals, and to homes in almost every part of daily life. In places where human intervention for quadcopter flight control is impossible, it becomes necessary to equip drones with intelligent autopilot systems so that they can make decisions on their own. All previous reinforcement learning (RL)-based efforts for quadcopter flight control in complex, dynamic, and unstructured environments remained unsuccessful during the training phase in avoiding the trend of catastrophic failures by naturally unstable quadcopters. In this work, we propose a complementary approach for quadcopter flight control using prediction error as an effective control policy reward in the sensory space instead of rewards from unstable action spaces alike in conventional RL approaches. The proposed predictive coding biased competition using divisive input modulation (PC/BC-DIM) neural network learns prediction error-based flight control policy without physically actuating quadcopter propellers, which ensures its safety during training. The proposed network learned flight control policy without any physical flights, which reduced the training time to almost zero. The simulation results showed that the trained agent reached the destination accurately. For 20 quadcopter flight trails, the average path deviation from the ground truth was 1.495 and the root mean square (RMS) of the goal reached 1.708.

Marine Autopilots’ Multipurpose Control Laws Synthesis for Actuators Time Delay

One analytical design problem involves constructing control laws for marine autopilot systems. Despite numerous known solutions, this problem can still be further developed by taking into account the actual conditions of the control system operation. An important issue for discussion is the feedback synthesis for marine ships with time delays in their rudders’ actuators. In this work, a new approach is proposed for providing all the desirable dynamic features of a closed-loop system with autopilot while taking into account the presence of a time delay. This approach is based on the predictive compensation of time delays via the specific transformation of an initially given reference controller with a special multipurpose structure. The applicability and effectiveness of the proposed method is further illustrated by a practical example of a controller design.

The Results of Field Test of the Mounted Attachment CHD-3

One of the ways of achieving landing accuracy is the control of the unit movement using navigator installed on the tractor, including autopilot systems that can correct the trajectory of the tractor. However, the straightness of tillage may be low. (Research purpose) The research purpose is in increasing the track stability of agricultural tools when cultivating row crops. (Materials and methods) The article proposes the use of a controlled attachment as part of a machine-tractor unit with navigation equipment installed on both a tractor and an agricultural tool, the use of which will significantly improve the accuracy of technological operations. The article presents the scheme, technical characteristics, and operating principle of the controlled attachment device. The article describes the method of conducting a field test of a machine-tractor unit with an installed controlled attachment. (Results and discussion) It was found that when cultivating without navigation equipment, the deviation from the axis was 62 mm; when cultivating with a machine-tractor unit with navigation equipment on a tractor it reaches 22 mm; when cultivating with a machine-tractor unit with a controlled CHD-3 attachment and navigation equipment on a tractor and cultivator it reaches 20 mm. The untreated area of the field by a machine-tractor unit with CHD-3 and navigation equipment on a tractor and cultivator was 0.27 square meters in relation to 0.45 square meters when processing the unit as part of a tractor with navigation equipment and a cultivator. (Conclusions) The use of tool control devices for cultivating rows of potatoes is impractical, since the straight movement of the planting machine along a given line is not guaranteed when planting potatoes. It was found that increased accuracy and further productivity will be achieved when using a controlled attachment on all operations of potato cultivation up to harvesting. The use of a controlled attachment device is advisable when cultivating crops that require high precision planting, for example, sugar beets.

Autonomous performance maximization of research-based hybrid unmanned aerial vehicle

Purpose This paper aims to investigate the autonomous performance optimization of a research-based hybrid unmanned aerial vehicle (i.e. HUAV) manufactured at Iskenderun Technical University. Design/methodology/approach To maximize the autonomous performance of this HUAV, longitudinal and lateral dynamics were initially obtained. Then, the optimum magnitudes of the autopilot system parameters were estimated by considering the vehicle’s dynamic model and autopilot parameters. Findings After determining the optimum values of the longitudinal and lateral autopilots, an improved design for the autonomously controlled (AC) HUAV was achieved in terms of real-time flight. Practical implications Simultaneous improvement of the longitudinal and lateral can be used for better HUAV operations. Originality/value In this paper, the autopilot systems (i.e. longitudinal and lateral) of an HUAV are for the first time simultaneously designed in the literature. This helps the simultaneous improvement of the longitudinal and lateral flight trajectory tracking performances.

Vision-based pose estimation of a multi-rotor unmanned aerial vehicle

Purpose The purpose of this paper is to facilitate autonomous landing of a multi-rotor unmanned aerial vehicle (UAV) on a moving/tilting platform using a robust vision-based approach. Design/methodology/approach Autonomous landing of a multi-rotor UAV on a moving or tilting platform of unknown orientation in a GPS-denied and vision-compromised environment presents a challenge to common autopilot systems. The paper proposes a robust visual data processing system based on targets’ Oriented FAST and Rotated BRIEF features to estimate the UAV’s three-dimensional pose in real time. Findings The system is able to visually locate and identify the unique landing platform based on a cooperative marker with an error rate of 1° or less for all roll, pitch and yaw angles. Practical implications The proposed vision-based system aims at on-board use and increased reliability without a significant change to the computational load of the UAV. Originality/value The simplicity of the training procedure gives the process the flexibility needed to use a marker of any unknown/irregular shape or dimension. The process can be easily tweaked to respond to different cooperative markers. The on-board computationally inexpensive process can be added to off-the-shelf autopilots.