The goal of this master project is to build an autonomous robot and explore the trade-off between gathering information regarding its physical surroundings and moving efficiently. We designed and built the robot with a fixed one-beam distance sensor. Thus the robot needs to rotate to scan the surrounding environment, which consumes both time and energy. We use omnidirectional wheels to decouple the rotation from the translation, so that the robot can drive in one direction with the translation, while measure the range in another direction with the rotation.
1. The one laser beam omnidirectional robot
To build the robot, shown in Figure 1, we augmented a basic omnidirectional robot kit with a laser distance sensor, a current and voltage sensor, an electronic board with an Inertial Measurement Unit and several analog inputs, and a microprocessor with wireless network.
Then, we used a simulator to develop the behaviour of the robot. The robot cannot measure the distance and the angle to the nearest wall directly, so we devised a behaviour that allows this to be estimated. The robot follows the wall in short, bouncing arcs to get more information about its position relative to the wall. Figure 2 displays this behaviour.
2. Bouncing behaviour of the robot
To explore the trade-off between sensing and moving, we vary the angle of rebound and compare the energy and the time spent to run through a lap. Shorter arcs mean less energy is used to gather information but result in poor quality wall following. Figure 3 presents the result of the experiment when comparing the time. The results of this experiment show that a large angle of rebound increases the efficiency of the robot.
3. Time for different angles of rebound
This thesis is a case study in autonomous mobile robotics where the results show that a robot with small sensing capability needs a lot of behaviour to achieve a task.
This project was done at the Autonomy Lab of Simon Fraser University, Vancouver in Canada,