Page 21 - Read Online

Page 466 Phadke et al. Intell Robot 2023;3:453-78 https://dx.doi.org/10.20517/ir.2023.27

learning approach called self-organizing maps. This methodology condenses the area-network coverage
problems into singular solutions that can output arrays of mobile agents over the target area. This broad
study can be programmed to military scenarios as well, where such efficient formations can provide
emergency communications to troops on the ground. The authors, however, have modeled this scenario
with SAR as their primary scenario descriptor.

Such methodologies can be used to complement existing swarm deployments using information exchange
policies for environmental awareness. Appearance-based tracking algorithms, such as , can detect victims
[77]
on the ground in SAR scenarios. It is possible to employ such techniques on collaborative UAVs to
effectively cover a larger area or track a single target from multiple frames to gain higher confidence levels
on target detection probability. Additional approaches involve using bimodal information-based target
[78]
[79]
recognition for victim detection . A collaborative process in uses a UAV-UGV to solve a SAR scene
locally in the absence of GNSS information. While GNSS is an integral part of robot navigation, it is also
prone to errors, signal sparsity, and jamming. Developing methodologies that can function without GNSS,
even temporarily, is an important addition to the overall resilience factor of a swarm. The UGV here is a
humanoid robot that localizes using a combination of local odometry and adaptive Monte Carlo
localization. In this study, a camera sensor backed by a neural network is used for detecting humans. The
aerial robot provides a 2.5-D map that is used as input by the path-planning process of ground robots. Such
hardware and operational space heterogeneous agents also highlight developments in the agent property
component. Decision independence in individual agents is a complex research task. It measures the
availability of decision-making capability that each agent is capable of during tasks. Agents that are capable
of making individual decisions have been shown to provide better results for SAR missions when compared
[80]
with a swarm that has centralized control . This is a problem that broadly falls under the resource
allocation and reassignment scheme for swarms. Agents are capable of requesting, holding on to, or
releasing common resources and reassignment tasks for accomplishing a common objective by making
individual decisions. The limitations to executing tasks by agents are their limited fuel and computing
capacity.

Path planning and obstacle avoidance fall under the control system area of swarm development using
optimization problems and environmental information based on simultaneously occurring external
incidents. Article uses multiple UAVs to decompose search grids and create efficient paths along polygon
[62]
edges in maritime SAR.

2.3. Target study and surveying
Target studying and surveying is a broad category that describes the use of UAV swarms for remote sensing
and ecological and agricultural scenarios. It differs from the above SAR section by its primary factor of time
constraints. While it is vital to accomplish victim detection and rescue in post-disaster scenarios, target
study, and surveying may require a long-term commitment to the interaction between the UAV swarm and
the target site. For example, following a herd of buffalos being hunted by lions to study pack hunting
strategies or using UAV swarms to create agricultural field vegetation maps for plant disease detection [22,81] .
[82]
Applications such as using fixed-wing aircraft to collect cumulus cloud data also fall under this category.
Table 5 highlights referred works that incorporate resilient mechanisms in target study applications.

Hence, there is an inherent shift in the way resiliency is perceived in such applications. A focus on sensor
data quality, sensor fusion, efficient transmission of data through network protocols, and energy-aware
routing protocols [83-88] is prevalent in resilient mechanisms. In persistent surveying applications such as
crowd control and surveying , there is no need for SAR protocols. However, UAV swarms may be used to
[89]