POTENTIAL IMPROVEMENTS

There are many improvements that you can make to our project. You can use the accelerometer data sent by the Robosapien’s Freescale node to predict which way the robot is moving relative to the accelerometers. This could be used to increase the accuracy of the probability function that evaluates samples in the particle filter. The result would be more accurate localization estimates. We successfully implemented this idea in the particle filter code, but it was impractical because of the Robosapien’s motion. When the Robosapien moves, it shifts its weight from side to side violently, making the accelerometer readings useless. For this method to contribute to the localization estimates, you’ll need a robot that moves more smoothly.

The best possible improvement, however, would be to add more ZigBee nodes. We didn’t have the resources to do so, but adding more nodes (to behave like the T node) would supply the localization routine with much more data. The result would be more accurate guesses as to robot’s location and thus better control over the robot.