Final Report

The goal of this project was to create a tour guiding robot by integrating several techniques and algorithms that would give tours of our science building as well as provide directions around the building. While such a robot may be viewed as a novelty, it's purpose was to generate more interest and awareness of the sciences at Bryn Mawr, especially Computer Science. We also wanted to gain more experience working with robotics since our prior experience was fairly limited.

In order to accomplish this large task, we started by researching previous robot projects to learn about which techniques and algorithms were used in projects similar to ours. We then focused our attention on creating a controller for the robot that would drive it down the center of the hallways in the building as smoothly as possible with collision detection. This task, although it deceptively seemed like the easiest task, proved to be non-trivial. However, once the robot had achieved the successful behavior, we progressed to work on localization.

To localize the robot, we came up with a strategy to use a combination of dead reckoning (which is notoriously inaccurate) and vision (which is expensive). We used the dead reckoning of the robot to approximate its location and then used vision to localize itself and recognize either a tour stopping / speaking landmark or a turning landmark in the tour which provided more precision to the task. A landmark consisted of a two inch strip of red or blue tape across the floor of the hallway. These landmarks were recognized by applying a color match filter which looked at each pixel in the captured image and see if it matched specific red, green, and blue values that were pre-set by us. The color match filter also examined the distance between the matching pixels. If the pixels were within a certain distance, they would be grouped resulting in a "blob" of the color that was being looked for. Once the blob size was a certain number of pixels, the landmark was considered to have been reached. At this point, the dead reckoning values would be reset to avoid accumulating a large amount of error in the odometry values. After completing this module, we continued with the directions module.

In the directions module, we used the well known A* algorithm to search a graph of the rooms in the building to find the shortest path and wrote an algorithm to convert the instructions into understandable English. We then integrated Sphinx developed by Carnegie Mellon University and Festival developed by The Centre for Speech Technology Research at the University of Edinburgh so that the robot could receive input from the users through its microphone and understand it as a series of commands and could verbalize instructions. We merged these modules into a single controller with use of a finite state machine. Our strategy during the research was to break down each problem into smaller, more manageable components. Some of these components were allocated to certain members of our team, and the rest were collaborations on our part.

The result of this project is a robot that is able to give a tour of part of our science building as well as provide directions for the rooms of one of the four floors. The robot gives a pre-scripted, pre-planned tour using dynamic navigation and localization. We hope that future Computer Science students at Bryn Mawr will expand this project by lengthening the tour or expanding the coverage of the directions module. The robot interacts with visitors using speech recognition with a limited vocabulary and gives directions verbally.

At this point the program used to control the robot is modular and easy to expand and enhance. The modularity, given by the Finite State Machine, enables other interested student to easily expand the project. This whole year of research taught us some important ideas. Now that we have completed the project, it seems rather easy to develop a tour guide robot. Looking back, we realize that it has been a long process of trial and error, that lead us to a successful end. After completing this research project, we feel that we have a better idea of what is necessary to do research and feel that with this experience we would be able to accomplish other projects with more ease.