AI/ML in UAV Networks

This work looks into the problem of a decentralized data offloading within an edge UAV swarm to mitigate the complexities of a single UAV continually generating and processing large application specific data. The mobile edge UAVs considered here are multi-rotor types having constrained energy and processing power, which makes long-term handling of large data volumes impossible for standalone UAVs. The load mitigation is carried out by offloading data from a source UAV to other swarm members with sufficient energy and processing requirements. In this work, we focus on selecting the most optimal multi-hop path through the UAVs concerning available energies and processing resources, which can survive the duration of the data offload between the source and a target UAV. We formulate a Multi-Armed Bandit (MAB) based offload path selection scheme, which selects the most energy and processing optimized multi-hop path between a source and a target UAV. Upon comparison of our scheme against the naive shortest path approach, we observe that our approach results in significant savings of collective network energies, even for long operational durations.

Read More PDF Citation

In this paper, we propose the use of a Long Short-Term Memory (LSTM) based server-side sequence prediction algorithm to ease network data-load caused by rapid polling of multiple sensors onboard aerial robotic platforms, which are wirelessly tethered to a remote server for control and coordination. Our scheme reduces the network access time latencies between these platforms and the remote server hosting the control and scheduling mechanisms. Reduction in the TDMA-based access time is achieved by reducing the actual amount of data transmitted over the network, using partial transmission of actual sensor data over the network and server-side sequence prediction of the voluntarily missed sensor values. Our scheme allows the TDMA control of an increased number of networked platforms without change of infrastructure or the network characteristics.

Read More PDF Citation
 

Smart UAV Networks in Agriculture

This work addresses the challenges of a decentralized and heterogeneous Unmanned Aerial Vehicle (UAV) swarm deployment -- some fitted with multimedia sensors, while others armed with scalar sensors -- in resource-constrained and challenging environments, typically associated with farming. Subsequently, we also address the resulting problem of sensing and processing resource-intensive data aerially within the Edge swarm in the fastest and most efficient manner possible. The heterogeneous nature of the Edge swarm results in under-utilization of the available computation resources due to unequal data generation within its members. To address this, we propose a Nash bargaining- based weighted intra-Edge processing offload scheme to mitigate the problem of heavy processing in some of the swarm members. We do this by distributing the data to be processed to all the swarm members. Real-life hardware tuned simulation of a large UAV swarm shows that by increasing the number of UAVs in the swarm, our scheme achieves better scalability and reduced processing delays for intensive processing tasks. Additionally, in comparison to regular star and mesh topologies, our scheme achieves an increase in collective available network processing speeds by 100% for only 25% of the number of UAVs in a star topology.

Read More PDF Citation

The gain in popularity of unmanned aerial vehicles (UAV), platforms and systems (UAS) can be attributed to its ease of operation, versatility and risk-free piloting. The primary UAV application domain has expanded, from recreational and military sights, to include scientific surveys and agriculture. The popularity of UAVs in scientific data gathering and applications, especially the use of small, multi-rotor UAVs is quite widespread. These multi-rotor UAVs are small, portable, low-cost, highly maneuverable, and easy to handle. These features make such UAVs attractive to scientists and researchers worldwide. There has been a sudden spurt of UAV use in niche domains, such as agriculture. Agriculturalists are choosing UAV-based field operations and remote sensing over the time-tested satellite-based ones, especially for local-scale and high spatiotemporal resolution imagery. In this survey, we explore various UAV application areas, types, sensors, research domains, deployment architectures. Comparisons between various UAV types, sensing technologies (UAV, WSN, satellites), UAV architectures and their utility in precision agriculture have been provided. Additionally, crop stress, its types, and detection using various remotely-sensed vegetation indices have been explored for their use in UAV-based remote sensing.

Read More PDF Citation
 

UAV Controls

We propose a method for a single ground camera-based visual gesture control of a quadrotor mUAV platform, making its flight more responsive and adaptive to its human controller as compared to a human controller using keypads or joysticks for controls. The proposed camera-based gesture control scheme provides an average accuracy of 100% gestures detected, as compared to accuracies obtained using expensive Kinectbased hardware, or processing intensive CNN-based pose estimation techniques with 97.5% and 83.3% average accuracies, respectively. A fog-based stabilization mechanism is additionally employed, which allows for flight-time stabilization of the mUAV, even in the presence of unbalanced payloads or unbalancing of the mUAV due to minor structural damages. This allows the use of the same mUAV without the need for frequent weight readjustments or mUAV calibration. This approach has been tested in real-time, both indoors as well as outdoors.

Read More PDF Citation