2019 International Conference on Intelligent Computing and Control Systems (ICCS) | 2019
A low power Artificial Intelligence Processor for Autonomous Mobile Robots
Abstract
The robot which makes use of AI as a mode of processing is getting more popular day by day, starting from the autonomous room cleaning robot to Amazon Prime Air. This autonomous robot overtakes traditional robots in following aspects such as implementing effective decision making in order to reduce the computational overhead by reducing the overall power usage of the robot. In this report, we have designed a low power [1] AIP without compensating in performance. The AIP which we have designed is a 64 processing element that uses parallel processing architecture. A map with 8 different routes is created in Xilinx where it calculates the shortest path from the source to destination using conditional operators. A* algorithm is implemented in Matlab to calculate the shortest distance and Dijkstra’s algorithm is converted to VHDL using Vivado HLS coder. A neural network is also created using Matlab to detect and avoid real time obstacle. The overall power report of the processor is implemented in Cadence.