"> Robot Test Bed

Robot Test Bed

NEW Gyro-Encoder Hybrid Localization

Using encoder data for localization is prone to positioning errors due to compounding of small angle errors. During a mapping, correct position knowledge is important and such errors can cause poor results; however, techniques such as registration and probabilistic positioning have been developed to cope with imperfect encoders. Improved low-level localization would relieve the robot of these expensive calculations, leaving more time for other tasks.

Encoders produce a fairly accurate measure of translation but poor measure of angle; gyroscopes measure angle highly accurately. Combining the devices by using the encoders for measuring distance and the gyro for measuring angle creates a more accurate and useful device. Presented are results comparing position calculations using solely encoders and combining encoders with a gyro. Test runs were carried out using a Active Media Pioneer II mobile robot and a KVH E-Core RD1100 rate gyro.

Odometry Experiments

     We are currently in the process of running experiments to quantify the error in the robot's odometry readings. The term "bias" refers to some constant multiple by which all the odometry readings are off; "drift" refers to the actual fluctuation in the robot's consistency of movement. In an ideal world, the robot would move exactly x meters if it was told to do so. But, due to idiosyncrasies in the robot's motor controller, inconsistencies in flooring, system delays, etc, there is almost always some X, Y, and Theta error. For example, when told to move forward twenty meters, it is possible that the robot's right wheel encoder moves slightly faster than the left, cause it to "drift" left as it progresses forward. In doing this, the robot's X position is affected, Y position is altered, and the Theta heading changes because the robot is essentially turning. Since the robots will be navigating by their internal odometry measurements, it is imper ative that we identify any drift or bias in the robot's movement to insure flawless navigation.

Experiment 1 - "Straight Line" Odometry Test

     This experiment was designed to test the X and Y drift of the robot. In each trial, the robot was started at the same X and Y position(imagine the starting point as (0,0) on a 2D coordinate plane). The robot was then issued a command to move forward either .5, 1, 2, 4, 7, or 10 meters (see diagram below). Upon completion of the movement, the actual X and Y distances the robot translated were measured. We then recorded the robot's internal X and Y odometry readings. This experiment was repeated eight times for each of the six possible distances. Ideally, the X distance reported and actual distance moved should be equal, and the Y distance and Theta heading should be 0.
     View Data and Results     Download Data and Results [30 KB]

Experiment 2 - "Around an Obstacle" Odometry Test

     This experiment was designed to test error in the angular heading of the robot. As above, the robot started at the same X, Y, and Theta positions(0,0,0). Also, an obstacle was placed approximately two meters directly in front of the robot. There were two sets of trials: first, the robot arched left around the obstacle, finishing at a position five meters forward in the X direction but with the same Y and Theta heading as the start position. Second, the robot arched right around the obstacle, once again finishing at a point five meters forward from where it started, but at the same Y and Theta heading as it started (see diagram below). Upon finishing movement, the robot's Theta odometry position was noted. We performed sixteen total trials, eight left and eight right. The reason we note a difference between left and right is in case there is a significant difference in the final readings reported in each condition. Ideally, the reported X position will equal five meters, and the Y and Theta heading should equal 0, for each condition. Upon completion of the experiment, we performed an Analysis of Variance (ANOVA) on the two conditions and found that there was no significant difference between left odometry readings and right odometry readings.
     View Data and Results     Download Data and Results [21 KB]
     View ANOVA Results       Download ANOVA Results [20 KB]

Experiment 3 - "Around the FRC" Odometry Test

    This experiment is similar to the above tests, only on a larger scale. The robot will once again start at position (0,0,0). The robot will then translate a path around the Field Robotics Center and end up in the same position it started. Therefore, ideal odometry readings from the robot would echo the start position - (0,0,0). As with the experiment above, we performed this experiment for two different conditions. In one condition, the robot moved clockwise around the FRC for 20 trials and in the other condition the robot moved counterclockwise, also for 20 trials. When we performed ANOVA on these results and found a significant difference between the clockwise and counterclockwise odometry readings for X, Y, and Theta. In addition, we performed ANOVA on the total distance error for the clockwise and counterclockwise conditions and found that difference to be significant as well. This means that we are getting consistently different values for the total distance error in the two co nditions.
     View Data and Results     Download Data and Results [18 KB]
     View ANOVA Results       Download ANOVA Results [23 KB]