SummaryLargeâscale wireless sensor networks (LSâWSNs) are used for collecting and monitoring the physical state of the environment. These networks are first sending information to the base station and then the recipient. LSâWSNs consist of several components, such as battery, sensor, transmitter, receiver and microcontroller circuits. If any one of these hardware components does not work properly, then the entire system will become faulty condition, resulting reducing network life and accuracy. To overcome these hardware node faults, this manuscript proposes an effective deep reinforcement learning (DRL)âadopted faulty node detection and recovery scheme (FNDâRS) integrated with the hosted cuckooâbased optimal routing scheme for LSâWSNs. Here, the DRL process is utilized for noticing and recovering the hardware node fault on sensor. The major objective of this system is âto enhance the network life with least energy consumption and also enhance the accuracy of Large Scale Wireless Sensor Networks using hosted cuckoo optimization (HOâCOA) algorithmsâ. The simulation is done in MATLAB. The experimental results shows that the proposed method attains lower overhead 15.65%, energy loss 21.62%, average packet delivery ratio 13.13%, energy efficiency 23.25%, network lifetime 19.12%, throughput 13.5%, time complexity 19.35% and lower delay 17.55% when compared to the existing methods, like FNDâRS for LSâWSN in distributed intermittent fault diagnosis algorithm (NDRâWSNâDISFD), FNDâRS for LSâWSN in harmony search algorithm (FNDRâWSNâHAS), FNDâRS for LSâWSN in tâdistributionâbased satin bowerbird optimization (FNDRâWSNâtâDSBO), FNDâRS for LSâWSN in optimal emperor penguin optimization (FNDRâWSNâOEPO) and FNDâRS for LSâWSN in particle swarm optimization algorithm (FNDRâWSNâPSO) respectively.