Project Title: Learning in Autonomous Robots using Artificial Neural Networks
This project will investigate the use of Artificial Neural Networks for learning in making robots autonomous. The project will use learning methods based on ANNs that will use sensory information from a robot in a controlled environment (e.g., a maze) to implement and verify the efficacy and extent of learning to enable a robot to be autonomous.
Aim: To apply Artificial Neural Networks (ANNs) for learning about the environment using sensory data from a roving robot.
Objectives: It is expected that the student will be able to
(i) Critically review the learning algorithms used in autonomous robotic applications;
(ii) Investigate the nature and characteristics of sensory data typically used in autonomous robotic applications;
(iii) Critically review and select one ANN architecture for investigation;
(iv) Implement and validate an algorithm for incremental learning using sensory data;
The project will develop and validate a Matlab-based suite (with C/C++/JAVA if necessary) that can take the sensory data from a roving robot and build a picture of the environment. The final stages of the project may involve compiling code to run independently on a programmable chip mounted on a robot.
1. Purchase of some necessary sensors and actuators for a robot may be necessary.
2. Matlab/Simulink software.