Pipelined, High Speed, Low Power Neural Network Controller for Autonomous Mobile Robot Navigation using FPGA
Najmuddin Aamer1, S. Ramachandran2
1Najmuddin Aamer, Research Scholar, Department of Electronics & Communication Engineering, Vinayaka Mission University, Salem (Tamil Nadu), India.
2S.Ramachandran, Department of Electronics & Communication Engineering, SJB Institute of Technology, Bangalore (Karnataka), India.
Manuscript received on 13 January 2016 | Revised Manuscript received on 22 January 2016 | Manuscript Published on 30 January 2016 | PP: 19-24 | Volume-5 Issue-8, January 2016 | Retrieval Number: H2259015816/16©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: The demand for autonomous robots which incorporates efficient path planning and obstacle avoidance is increasing rapidly. In this paper, we have proposed a neural network based hardware architecture for autonomous mobile robot which is able to detect and avoid obstacles by using prediction model of neural network and distribution computation techniques using FPGA. Learning and prediction is implemented by using back propagation method on FPGA Virtex-II pro kit. For flexibility and accuracy of the neural network, floating point based computation method is applied. The proposed model uses the principle of reconfigurability which reduces the implementation cost and area. In this proposed architecture of autonomous mobile robot, pipelined architecture is used which increases the speed and reduces the delay for the prediction. Simulation is performed by using Xilinx 14.3 ISE simulator. Place and Route results exhibit high throughput and low power consumption achieved using this proposed model for controlling the autonomous robot.
Keywords: Autonomous Mobile Robot, FPGA, Neural Network, Pipeline, Reconfigurability, Path Planning and Obstacle Avoidance.
Scope of the Article: Neural Information Processing