We consider a network of unmanned aerial vehicles (UAVs) for a search-and-rescue operations involving both detection of multiple targets and mapping of environment, where the learning time is limited. One possibility for accomplishing the goal while guaranteeing short learning time is to employ cooperation among UAVs. With this objective, we adopt a multi-agent Q-learning algorithm that allows the UAVs to learn a suitable navigation policy in real-time in order to complete a mission within a fixed time frame. The obtained results demonstrate that proper combination of the information gathered by the UAVs allows for an accelerated learning process.
Multi-Agent Q-Learning in UAV Networks for Target Detection and Indoor Mapping / Guerra A.; Guidi F.; Dardari D.; Djuric P.M.. - ELETTRONICO. - (2021), pp. 9593232.80-9593232.84. (Intervento presentato al convegno 4th International Balkan Conference on Communications and Networking, BalkanCom 2021 tenutosi a Serbia nel 20-22 September 2021) [10.1109/BalkanCom53780.2021.9593232].
Multi-Agent Q-Learning in UAV Networks for Target Detection and Indoor Mapping
Guerra A.
;Guidi F.;Dardari D.;
2021
Abstract
We consider a network of unmanned aerial vehicles (UAVs) for a search-and-rescue operations involving both detection of multiple targets and mapping of environment, where the learning time is limited. One possibility for accomplishing the goal while guaranteeing short learning time is to employ cooperation among UAVs. With this objective, we adopt a multi-agent Q-learning algorithm that allows the UAVs to learn a suitable navigation policy in real-time in order to complete a mission within a fixed time frame. The obtained results demonstrate that proper combination of the information gathered by the UAVs allows for an accelerated learning process.File | Dimensione | Formato | |
---|---|---|---|
PP Multi-Agent Q-Learning.pdf
Open Access dal 14/05/2022
Tipo:
Postprint
Licenza:
Licenza per accesso libero gratuito
Dimensione
480.67 kB
Formato
Adobe PDF
|
480.67 kB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.