A Novel Deep Q-Reinforcement Learning Model for building Efficient Agriculture Autonomous Robots

Prashanth M V,Vijaya Kumar M V

digital farming is the practice widely adopted in modern technologies such as sensor, data analytics, and robotics for moving from tedious process to more continuous automated processes. Firstly, this work present a deep rooted survey in agricultural robotics for performing various task such as soughing, harvesting etc. From survey it can be seen adopting industrial robots for agricultural purpose is not a viable option due to complex environment. For understanding the unknown environment cognitive architecture has been applied. CA are used to model artificial minds capable of behaving like human (farmer). Recently, Society of Mind Cognitive Architecture (SMCA) has been presented using multi-agent and reinforcement learning technique. However, it is problematic to solve the Markov decision process problem because the Markov decision process has many variables. Further, no prior work has considered such cognitive architecture design for agriculture domain. For overcoming research challenges, this paper presents a Deep Q- Reinforcement Learning (DQRL) cognitive learning technique for SMCA by combining both reinforcement learning and Deep learning method. The DQRL method is used to control the communication power of agents/robots based on dynamic environment requirement. Experiment outcome shows DQRL attain. superior performance than existing model in terms of energy efficiency, learning efficiency and memory efficiency.

Volume 11 | 02-Special Issue

Pages: 587-594