Page 27 - Read Online
P. 27

Page 472                         Phadke et al. Intell Robot 2023;3:453-78  https://dx.doi.org/10.20517/ir.2023.27

                              [123]
               their applications .

               There are several directions that justify further exploration. A multi-objective optimization approach would
               hold promise for balancing diverse objectives while enhancing resilience. Dynamic adaptation mechanisms,
               powered by machine learning and AI, can facilitate real-time adjustments based on evolving conditions.
               Additionally, fostering effective human-swarm interaction techniques and exploring innovative sensor
               configurations can amplify the resilience of UAV swarms.


               The significance of our findings resonates in the advancement of resilient UAV swarm applications across
                                                               [124]
               various domains. From disaster response to agriculture , the potential impact on societal well-being is
               substantial. This survey is designed to motivate readers to contemplate the intricate dynamics of UAV
               swarm resilience, to critically assess their applicability within their fields, and to contribute to the ongoing
               discourse in this area of research.

               5. FUTURE RESEARCH DIRECTIONS
               Few researchers have addressed the problem of swarm agent resilience and well-being as they perform their
               assigned tasks. Depending on the resilience thresholds designed, the loss of several agents during mission
               progress can affect the ability of the swarm to complete its mission. Even with tight constraints, some agents
               can exhibit degraded performance and fail. As such, the swarm must have the capability for self-awareness
               of the location and well-being of its agents. Previous studies have not addressed methodologies to track and
               rescue their agents in the case that a loss occurs. SAR of swarm agents has been recognized as component
                   [15]
               three  by the authors of this study, where we found that there was almost no current research to reference
               its implementation. One of the closest approaches to awareness policy by swarms to replace lost UAVs is
               described in , which uses a replacement policy to replace lost UAVs. Recovery policies are almost absent
                          [125]
               in current deployments where UAV swarm actively tries to recover lost agents. A different approach that
               the authors of  took was to design rescue depots and dynamic mission abort policies for swarm agent
                           [126]
               well-being. These are perfect examples of preliminary work on SS-SAR procedures.

               The presence of heterogeneity in swarms can be categorized by a variety of factors, such as their hardware
               buildup, operational space, and agent property. Operational space heterogeneity occurs when a swarm
               comprises of agents working across a diverse target space, such as quadcopters and water surface vehicles
               or ground-based  rovers.  Heterogeneity  by  nature  is  introduced  when  heterogeneity  is  induced  by  the
               assigned operational characteristics in homogeneous hardware agents, such as a fast and slow agent or
               exploratory and cautious agent combination swarms. Although challenging to address, heterogeneous
               swarms  have been  observed  to  produce  better  performance,  including  co-evolution  and  the  natural
               emergence  of  agent capabilities [127,128] .  This  occurs  due  to  complementary  abilities  brought  about  by
               heterogeneous  agents and extended thresholds on agent ability. Multiple demonstrations of operational
               space combinations have been trialed, such as air-underwater vehicles , air-water surface , and air-
                                                                             [129]
                                                                                               [130]
                             [25]
               ground vehicles . A selection of research exists that specifically targets heterogeneous agent issues in
               swarms  as  discussed  above  in  each category;  however,  prevalent  issues  with  implementation  along
               with wide-ranging assumptions during experimental designs warrant further research.

               6. CONCLUSION
               This study conducts a general review of different application scenarios and the various novel resiliency
               mechanisms that have been proposed to make swarms more efficient in accomplishing assigned goals. The
               end goal of systemic resiliency research is the complete integration of such mechanisms in every facet of
               system operation. This is an ambitious goal by itself, which requires ground-up development for all system
   22   23   24   25   26   27   28   29   30   31   32