Fig 2 - uploaded by Matt Garratt
Content may be subject to copyright.
Source publication
This paper describes a new non-linear control technique applied to the heave control of an unmanned rotorcraft. First a hybrid plant model consisting of exactly known dynamics is combined with a black-box representation of the unknown dynamics. Desired trajectories are calculated to smoothly achieve a sequence of random step changes in desired heig...
Context in source publication
Context 1
... are controlled using Pulse Width Modulation (PWM) at an update rate of 50 Hz. The PWM signals sent to the servo are logged in units of microseconds (μs) corre- sponding to the width of the control pulse sent to each servo. The MPC555 autopilot generates the PWM servo signals for the five servo channels. The Eagle avionics system is shown in Fig. ...
Similar publications
Combining deep neural networks with reinforcement learning has shown great potential in the next-generation intelligent control. However, there are challenges in terms of safety and cost in practical applications. In this paper, we propose the Intervention Aided Reinforcement Learning (IARL) framework, which utilizes human intervened robot-environm...
Advantages in the application of intelligent approaches, such as the conjunction of artificial vision and Unmanned Aerial Vehicles (UAV), have been recently emerging. This paper presents a system capable of detecting ground vehicles through aerial images taken by a UAV in real time. In addition, the system offers the possibility to autonomously gui...
Citations
... The adverse cushioning effects can be regarded as continuous stochastic disturbances during the flights. Accordingly, the relationship between thrust and control input can be altered by up to 50 %, increasing the non-linearity of the system dynamics (Garratt and Anavatti, 2012). ...
Robotic aircraft are often required to operate in harsh environments (e.g., underground mining, cluttered environments, and battlefields). In this chapter, we discuss an adaptive (evolving) fuzzy system that has the ability to learn and to configure itself based on the human way of learning, which is also somewhat akin to the principles of natural evolution. We will be looking at the capability of an evolving Takagi-Sugeno (ETS) fuzzy algorithm to learn-from-scratch in order to adapt the challenging dynamics of autonomous systems in real-time. The ETS system can also work in unknown environments, where there is no expert knowledge. While we focus on the implementation of the ETS system to identify the behavior of a fast-dynamical system as in the case of the low altitude hovering of our Tarot hexacopter drone by performing an online ETS-based data driven modelling (online system identification) technique, we also conduct a preliminary study to highlight the efficacy of the ETS autopilot under computer simulations.
... The primary contribution of this paper is the successful development of spiking neurocontrol for a simulated hexacopter, which is accomplished by decomposing it into modular networks and evolving them incrementally with the MoNEAT algorithm. Along with our previous work, we have created a pathway to developing flight control for physical UAV platforms, by beginning from identification of the system [28], to decomposing the controller into modules (in this paper), and finally to integrating Hebbian plasticity for offline-online hybrid training [23]. ...
... In [58] (2012), an offline feed-forward network is trained for learning heave (or vertical) dynamics of an unmanned helicopter. The network has a single hidden layer consisting of 4 hidden neurons. ...
... In [58] (2012), a two-layer feed-forward network with 4 hidden neurons is trained offline using LM algorithm to mimic the control inputs generated from an optimal feedback control law for the heave control of a helicopter. Both simulation and flight test results show satisfactory tracking performance. ...
Model-Based Control (MBC) techniques have dominated flight controller designs for Unmanned Aerial Vehicles (UAVs). Despite their success, MBC-based designs rely heavily on the accuracy of the mathematical model of the real plant and they suffer from the explosion of complexity problem. These two challenges may be mitigated by Artificial Neural Networks (ANNs) that have been widely studied due to their unique features and advantages in system identification and controller design. Viewed from this perspective, this survey provides a comprehensive literature review on combined MBC-ANN techniques that are suitable for UAV flight control, i.e., low-level control. The objective is to pave the way and establish a foundation for efficient controller designs with performance guarantees. A reference template is used throughout the survey as a common basis for comparative studies to fairly determine capabilities and limitations of existing research. The end-result offers supported information for advantages, disadvantages and applicability of a family of relevant controllers to UAV prototypes.
... Also, rotary wing aircraft generally operate at lower speeds and altitudes compared to fixed wing aircraft, making them suitable for challenging tasks, such as building and infrastructure inspections, even for indoor tasks such as large aircraft inspections inside a hangar. Accordingly, when flying at low altitudes, the system is prone to the adverse impacts of ground effect, which can introduce significant nonlinearity by modifying the relationship between thrust and control input by up to 50 % [7]. ...
... Consequently, while one can employ linear mathematical models, there are many unmodeled dynamics that cannot be fully captured using the concept of transfer functions. For instance, in the heave (vertical) dynamics of a helicopter, there are several non-linear dynamics that can cause uncertainties, such as wind gusts, kinematics of actuators and servo dynamics, variations in rotor speed as well as sensor lag, and ground effects [7]. Nonetheless, since it is not trivial to model those uncertainties, many research papers in the literature have not properly taken them into account, resulting in overly simplified dynamic models. ...
... where N denotes the number of fuzzy rules and M highlights the number of inputs. The free parameters of the fuzzy model in (7), updated in the adaptation process, are c ji , σ ji , b j , and σ i , indicating the center and the spread of the input/output fuzzy membership functions. 1) Back Propagation of Error: As part of its online learning procedure, our generated model is connected in parallel with the plant to identify its online behavior (e.g., the consistency and the accuracy of the system in the face of uncertainties). ...
This article studies nonlinear system identification of a small scale and flybar-free unmanned helicopter, the Trex450 chopper, built using commercial off-the-shelf components. We employ the real-time input-output data, obtained from human-controlled flight tests, operating the aircraft under severe ground effects during the vertical flight maneuvers. We highlight the efficacy of the entropy fuzzy system identification method with respect to the performance of several well-known nonlinear system identification techniques (i.e., a Takagi-Sugeno Fuzzy system, an adaptive neuro-fuzzy inference system (ANFIS), and a nonlinear autoregressive with exogenous (NARX) model) as our benchmarks. Our research confirms the benefits of the entropy fuzzy identification technique. Despite being nonlinear, the proposed fuzzy model is relatively simple, transparent, and highly accurate to represent the complex non-linear dynamic behaviors of our unmanned helicopter under severe ground effects. Another major advantage of the proposed system identification technique is its ability to avoid overfitting, an essential requirement in modeling. Overall, the fuzzy system is also capable of achieving a delicate balance between maximizing the accuracy while minimizing the complexity of the acquired model.
... Not focusing on landing specifically, Razmi and Afshinfar [24] developed a neural network based control system for position and attitude tracking control of a UAV. Focusing on the landing, Garratt and Anavatti [25] used a neural network controller to produce heave trajectories for a UAV. Similarly, Moriarty et al. [16] suggest using a stereo camera to track a ship's landing zone and a neural network to predict opportune landing windows. ...
... (20) and (22) respectively, while case 2 and 4 both used Eq. (25) and (24). Table 2 summarizes the landing velocity results obtained from the tests. ...
One of the key challenges to the safe operation of UAVs at sea is the relative motion that exists between the UAV and ship during landing. The scope of this work is the development and evaluation of methodologies for improving UAV landing performance. The new methodologies are known as the Signal Prediction Algorithm (SPA), Active Heave Compensation (AHC) and the Landing Period Indicator (LPI). To promote interoperability, the methodologies do not require any specific ship equipment.
To evaluate the methodologies on an existing comprehensive UAV model, ShipMo3D was used to generate 105 sets of ship motion in sea states 2–6, headings 0–180∘ and speeds 6–10 knots. Within the simulation the UAV, equipped with a Light Detection and Ranging (LIDAR) system, measures the ship motion in situ. Using the Signal Prediction Algorithm (SPA), the UAV forecasts the ship motion and potential landing opportunities. The UAV can also use the SPA and the AHC system to maintain a safe hover position above the ship deck and determine landing trajectories with a specific touchdown velocity. The UAV can also employ the novel Landing Period Indicator (LPI) system which calculates an estimate of the ship's energy and determines opportune times for safe landings.
The results indicate that the methodologies can improve the landing performance of autonomous helicopters. For the 105 sets of ship motion, using the combination of the SPA, AHC, and LPI improved landing success by up to 34% when compared to a common landing controller.
... When simulating, it is not uncommon to derive UAV models mathematically from first principles [2,19]. However, such models are ill-suited to capturing every aspect of the system dynamics, because some of them cannot easily be modeled analytically, e.g., actuator kinematic nonlinearities, servo dynamics, etc [6]. Ignoring these effects can significantly deteriorate the performance of the designed controller when being deployed onto the targeted platform. ...
... This identified model is obtained by applying a data-driven process called 'system identification' that models the exact dynamics from the measured plant's input and output data. Such implementations have been successful amongst previous research [6,9,14,17,18]. ...
A critical issue in evolutionary robotics is the transfer of controllers learned in simulation to reality. This is especially the case for small Unmanned Aerial Vehicles (UAVs), as the platforms are highly dynamic and susceptible to breakage. Previous approaches often require simulation models with a high level of accuracy, otherwise significant errors may arise when the well-designed controller is being deployed onto the targeted platform. Here we try to overcome the transfer problem from a different perspective, by designing a spiking neurocontroller which uses synaptic plasticity to cross the reality gap via online adaptation. Through a set of experiments we show that the evolved plastic spiking controller can maintain its functionality by self-adapting to model changes that take place after evolutionary training, and consequently exhibit better performance than its non-plastic counterpart.
... Lee et al. [143,144] discussed the back-stepping control technique for a small-scale unmanned helicopter under varying load conditions and obtained better control performance than [137]. Fang et al. [145] and Sun et al. [146] In addition, some advances have been made in the application of intelligent and nonlinear control methods such as fuzzy logic control [153,154], neural network control [155][156][157][158][159], trajectory linearization [112], dynamic inversion [160]and adaptive control [161] to autonomous flight control of the small-scale unmanned helicopter. Back-stepping [162,163] and sliding mode controller [164] design based on the linearized model are applied to the miniature unmanned helicopter (or nonlinear model) and showed satisfactory tracking performance. ...
Miniature helicopters are highly unstable, agile, nonlinear under-actuated systems with significant inter-axis dynamic coupling. They are considered to be much more unstable than fixed-wing unmanned air vehicles, and constant control action is required at all times. However, helicopters are highly flexible aircraft, having the ability to hover, maneuver accurately and carry heavy loads relative to their own weight. Fixed-wing aircraft are used for application in favorable non-hostile conditions but in adverse conditions, agile miniature helicopters become a necessity. The conditions where a helicopter can perform better than fixed-wing UAVs include military investigation, bad weather, firefighting, search and rescue, accessing remote locations and ship operations. In such conditions, helicopters are subjected to unknown external disturbances such as wind and ground effect. These external disturbances have a significant opposing effect on the helicopter stability and can have disastrous results in extreme cases. So, designing and development of a controller for the miniature unmanned helicopter which can effectively reject the effect of these unknown external disturbances are of great theoretical significance and practical utility.
The principal part of this dissertation is the development of a simplified nonlinear model of the miniature unmanned helicopter. Then research on the development of the robust nonlinear control techniques for it based on the linearized model of the miniature helicopter.
Firstly, a simplified 11th order nonlinear model of the miniature unmanned helicopter is derived based on the previously published research work in international journals. Then the model is linearized at hover condition and divided into two separate subsystems (longitudinal-lateral and heading-heave subsystem). At hover condition, these two subsystems are weakly coupled and are suitable for separate controller design which makes the controller design procedure very simple. But during forward flight and aggressive maneuvers helicopter states are far from hover condition which deteriorates its control performance. Disturbance observer is used for this model mismatch and external disturbances approximation. Combining disturbance observer with sliding mode control, superior tracking performances are achieved for the miniature unmanned helicopter.
The core contributions and main innovations of this dissertation are the following.
1. Disturbance observer based sliding mode control for the small-scale unmanned helicopter hover operation.
2. Fixed time disturbance observer based fixed time sliding mode control for the small scale unmanned helicopter hover operation.
3. Extended state disturbance observer based sliding mode control for the small-scale unmanned helicopter path tracking in complex wind conditions
... In [58] (2012), an offline feed-forward network is trained for learning heave (or vertical) dynamics of an unmanned helicopter. The network has a single hidden layer consisting of 4 hidden neurons. ...
... In [58] (2012), a two-layer feed-forward network with 4 hidden neurons is trained offline using LM algorithm to mimic the control inputs generated from an optimal feedback control law for the heave control of a helicopter. Both simulation and flight test results show satisfactory tracking performance. ...
... Researchers have examined methods that allow for autonomous flight control, and landings on both stationary and moving decks. Garratt and Anavatti [8] used a neural network controller to produce heave trajectories for an unmanned helicopter. Hervas et al. [9] developed a landing control algorithm for unmanned vehicles on moving platforms that controlled the landing based solely on the relative heave motion between UAV and ship deck. ...
... Garratt and Anavatti [90] proposed a non-linear control technique for controlling the heave of an unmanned rotorcraft as seen in Fig. 3a. The authors developed a hybrid plant model, comprised of the known dynamics and a black box representation of it. ...
... This also eliminates the need for modelling and system identification technique. Neural network control of the heave dynamic of an unamnned helicopter according to [90]: (a) Eagle helicopter platform, (b) Non-linear characteristics of the collective with respect to vertical velocity and acceleration, (c) Performance of the neural network control system vs. conventional PD control. ...
We discuss state-of-the-art intelligent robotic aircraft with the special focus on evolutionary autopilots for small unmanned aerial vehicles (UAVs). Under the umbrella of adaptive autopilots, we highlight the pros and cons of the most widely implemented intelligent algorithms against the navigational and maneuvering capabilities of small UAVs. We present several cutting-edge applications of bioinspired flight control systems that have the capability of self-learning. We also highlight several research opportunities and challenges associated with each technique.
