An examination of trends in the growing scientific literature on approaching wildlife with drones

Drones or unoccupied aerial vehicles are rapidly being used for a spectrum of applications, including replacing traditional occupied aircraft as a means of approaching wildlife from the air. Though less intrusive to wildlife than occupied aircraft, drones can still cause varying levels of disturbance. Policies and protocols to guide lowest-impact drone flights are most likely to succeed if considerations are derived from knowledge from scientific literature. This study examines trends in the scientific literature on using drones to approach wildlife between 2000 and 2020, specifically in relation to the type of publications, scientific journals works are published in, the purposes of drone flights reported, taxa studied, and locations of studies. From 223 publications, we observed a large increase in relevant scientific literature, the majority of which were peer-reviewed articles published across 87 scientific journals. The largest proportions of peer-reviewed research articles related to aquatic mammals or aquatic birds, and the use or trial of drone flights for conducting population surveys, animal detection or investigations of animal responses to drone flights. The largest proportion of articles were studies conducted in North America and Australia. Since animal responses to drone flights vary between taxa, populations, and geographic locations, we encourage further growth in the volume of relevant scientific literature needed to inform policies and protocols for specific taxa and/or locations, particularly where knowledge gaps exist.

The utility of drones for studying polar bear behaviour in the Canadian Arctic: opportunities and recommendations

Climate-induced sea-ice loss represents the greatest threat to polar bears (Ursus maritimus), and utilizing drones to characterize behavioural responses to sea-ice loss is valuable to forecasting polar bear persistence. In this manuscript, we review previously published literature and draw on our own experience of using multirotor aerial drones to study polar bear behaviour in the Canadian Arctic. Specifically, we suggest that drones can minimize human-bear conflicts by allowing users to observe bears from a safe vantage point; produce high-quality behavioural data that can be reviewed as many times as needed and shared with multiple stakeholders; and foster knowledge generation through co-production with northern communities. We posit that in some instances drones may be considered as an alternative tool for studying polar bear foraging behaviour, interspecific interactions, human-bear interactions, human safety and conflict mitigation, and den-site location at individual-level, small spatial scales. Finally, we discuss flying techniques to ensure ethical operation around polar bears, regulatory requirements to consider, and recommend that future research focus on understanding polar bears’ behavioural and physiological responses to drones and the efficacy of drones as a deterrent tool for safety purposes.

Detection and tracking of belugas, kayaks and motorized boats in drone video using deep learning

Aerial imagery surveys are commonly used in marine mammal research to determine population size, distribution and habitat use. Analysis of aerial photos involves hours of manually identifying individuals present in each image and converting raw counts into useable biological statistics. Our research proposes the use of deep learning algorithms to increase the efficiency of the marine mammal research workflow. To test the feasibility of this proposal, the existing YOLOv4 convolutional neural network model was trained to detect belugas, kayaks and motorized boats in oblique drone imagery, collected from a stationary tethered system. Automated computer-based object detection achieved the following precision and recall, respectively, for each class: beluga = 74%/72%; boat = 97%/99%; and kayak = 96%/96%. We then tested the performance of computer vision tracking of belugas and manned watercraft in drone videos using the DeepSORT tracking algorithm, which achieved a multiple-object tracking accuracy (MOTA) ranging from 37% – 88% and multiple object tracking precision (MOTP) between 63% – 86%. Results from this research indicate that deep learning technology can detect and track features more consistently than human annotators, allowing for larger datasets to be processed within a fraction of the time while avoiding discrepancies introduced by labeling fatigue or multiple human annotators.

Potential Cyber Threats, Vulnerabilities, and Protections of Unmanned Vehicles

This study seeks to contribute to the literature by presenting a discussion of potential cyber risks and precautionary measures concerning unmanned vehicles as a whole. In this study, Global Navigation Satellite System (GNSS) spoofing, jamming, password cracking, Denial-of-Service (DoS), injecting malware, and modification of firmware are identified as potential cyberattack methods against unmanned vehicles. Potential deterrents against the aforementioned cyberattack methods are suggested as well. Illustrations of such safeguards include creating an architecture of the multi-agent system, using solid-state storage components, applying distributed programming tools and techniques, implementing sophisticated encryption techniques for data storage and transmission, deploying additional sensors and systems, and comparing the data received from different sensors.

Pilots’ Willingness to Operate in Urban Air Mobility Integrated Airspace: A Moderated Mediation Analysis

Interest in advanced air mobility (AAM) and urban air mobility (UAM) operations for on-demand passenger and cargo transport continues to grow. There is ongoing research on market demand and forecast, community acceptance, privacy, and security. There is also ongoing research by NASA, FAA, academia, and industry on airspace integration, regulatory, process, and procedural challenges. Safe integration of UAM and AAM will also require different stakeholder perspectives such as air traffic controllers, manned aircraft pilots, remote pilots, UAM operators, and the community. This research aimed to assess the willingness of manned aircraft pilots to operate in UAM integrated airspace based on airspace complexity and UAM automation level. In addition, a moderated mediation analysis was conducted using trust and perceived risk as mediators and operator type as a moderating variable. The results indicated that automation level influenced pilots’ willingness to operate an aircraft in integrated airspace. A moderating effect of operation type on automation level and willingness to pilot an aircraft was also observed: professional pilots were more amenable to UAM operations with a pilot on-board compared to remotely piloted operations. Results from the study are expected to inform airspace integration challenges, processes, and procedures for UAM integrated operations.

Adaptive Augmented Torque Control of a Quadcopter with an Aerial Manipulator

A quadcopter manipulator system is an aerial robot consisting of a quadcopter with a robotic arm attached to it. The system has coupled non-linear dynamics with uncertain time-varying parameters. The work in this paper focuses on designing an adaptive non-linear controller to facilitate the uncertain system’s trajectory tracking and stability. The novelty of the proposed work is the design and implementation of an adaptive feedback linearization controller, called adaptive augmented torque (AAT) control, for the aerial robot. The control law is based on a feedback linearization controller with model reference adaptive controller and a tracking error-based augmented term. Using the input-to-state (ISS) stability concept, a bound on the parameter estimation error is also developed. In the presented methodology, the controller uses estimated values of system parameters obtained from the adaptive mechanism and the tracking error to compute the control input using the AAT control law. An adaptive law for estimating unknown parameters is obtained using the strictly positive real-Lyapunov method. The asymptotic stability of the closed-loop system is analyzed via the Lyapunov theory. Simulations implemented in MATLAB and ROS/Gazebo and preliminary hardware experiments are presented to validate the theoretical results and to corroborate the performance of the AAT control law.

A New Scoring System for use in Capture-Recapture Studies for Bowhead Whales Photographed with Drones

Effective management of animal populations requires knowledge of life history parameters and estimates of population abundance. One method commonly used to estimate abundance is capture/recapture analyses of photographs. Small, relatively inexpensive, rotary-wing drones have become an effective platform for obtaining high-quality aerial photographs of whales. To conduct capture/recapture analyses the animal needs to be defined as marked or unmarked and the photographs must be of high quality. While a system for scoring quality and markedness has previously been developed for bowhead whales (Balaena mysticetus) (Rugh et al. 1998), a revised scoring system was needed to incorporate increased information in photographs taken by drones. We present a revised scoring system that enlarges two of the previously defined areas of the whale examined for markings and incorporates smaller markings into the definition of marked whales. We scored 30 whales using the previous criteria and the revised criteria developed in this paper. More whales were identified as marked (23%) and mark scores were higher for 30% of the zones scored using the new system. Increasing the number of marked whales during capture/recapture studies increases the precision of estimated parameters and permits us to make those estimates with smaller samples of photographs.

Nesting common eiders (Somateria mollissima) show little behavioral response to fixed-wing drone surveys

Drones may be valuable in polar research because they can minimize researcher activity and overcome logistic, financial, and safety obstacles associated with wildlife research in Polar Regions. Because Polar species may be particularly sensitive to disturbance and some research suggests behavioral responses to drones are species-specific, there is a need for focal species-specific disturbance assessments. We evaluated behavioral responses of nesting Common Eiders (Somateria mollissima, n =19 incubating females) to first, second, or in a few cases third exposure of fixed-wing drone surveys using nest cameras. We found no effect of drone flights (F1,23 = 0, P < 1.0) or previous exposures (F1,23 = 0.75, P = 0.397) on the probability of a daily recess event (bird leaves nests). Drone flights did not impact recess length (F1,25 = 1.34, P = 0.26); however, eiders with prior drone exposure took longer recess events (F1,25 = 5.27, P = 0.03). We did not observe any overhead vigilance behaviors common in other species while the drone was in the air, which may reflect eider’s anti-predator strategies of reducing activity at nests in response to aerial predators. Surveying nesting common eider colonies with a fixed-wing drone did not result in biologically meaningful behavioral changes, providing a potential tool for research and monitoring this Polar nesting species.

Quadrotor Motion Control Using Deep Reinforcement Learning

We present a deep neural net-based controller trained by a model-free reinforcement learning (RL) algorithm to achieve hover stabilization for a quadrotor unmanned aerial vehicle (UAV). With RL, two neural nets are trained. One neural net is used as a stochastic controller which gives the distribution of control inputs. The other maps the UAV state to a scalar which estimates the reward of the controller. A proximal policy optimization (PPO) method, which is an actor-critic policy gradient approach, is used to train the neural nets. Simulation results show that the trained controller achieves a comparable level of performance to a manually-tuned PID controller, despite not depending on any model information. The paper considers different choices of reward function and their influence on controller performance.