Intelligent micro-navigation system for autonomous robots

Autonomous vehicle or self-driving vehicle prototypes are being developed rapidly and are soon expected to become widespread. One can foresee a prospective future with ease of travel, fewer traffic collisions, congestions and higher travel speeds. Indoor autonomous vehicles possess applications such as firefighting, military combat, industrial-leak detection, etc. Prototypes being developed today prominently utilize a combination of GPS (Global Positioning System) data and IMU (Inertial Measurement Unit) data. However GPS data is not ubiquitous especially in car-parks, tunnels, within buildings, etc. Further, their localization accuracy has an error radius of 3 - 6m [1] which is unsuitable for indoor navigation. Thus to overcome this issue, we envisioned substituting the use of GPS with passive RFID tags. These tags would act as microsatellites, conveying their location to the vehicle, upon request. Thus autonomous vehicles installed with a RFID Reader would be able to read the tags and interpret their location. With this motivation, an autonomous robot was theoretically modeled in MATLAB R2012a and algorithms for localization, vehicle control and obstacle avoidance were successfully formulated and verified. Extended Kalman Filtering was incorporated to improve the localization. Further, to validate these algorithms and test their accuracy for indoor localization, a robot platform was developed. This platform was interfaced with Ultrasonic sensors for obstacle avoidance, IMU for heading estimation, RFID Reader to read RFID tags and XBee for wireless transmission of data. The localization was tested and found to have an error radius of < 10 cm. This is a significant improvement from the GPS based localization which has 3 – 6m error radius. To extend the application of this micro-navigation system to a swarm of robots, a novel Dynamic Task Allocation algorithm was formulated and simulated in MATLAB R2012a. This algorithm incorporates key features such as context awareness and team based task allocation. A Monte-Carlo simulation was performed to assess the algorithm and to compare it with the popular Random Choice Algorithm [14] for dynamic task allocation. The simulation showed the algorithm’s consistency in achieving the team’s objective and was found to be significantly better than the Random Choice Algorithm. In the future, it is intended to implement this algorithm on a swarm of micro-navigation systems for industrial, military, firefighting and other applications.Bachelor of Engineerin