How many years have you been competing in in RCJ Soccer?
Three years
What’s the most interesting thing you’ve prepared for this year’s RoboCupJunior Soccer competition?
The most interesting thing that we added to our robot is the new Arduino 33 BLE Sense, besides having a powerful microprocessor, it has a BLE communication module in the board and other sensors, like the 9-axis IMU, which we use in the robot localization. Here is a link with the board specifications: https://store.arduino.cc/usa/nano-33-ble-sense
It looks like you are on a good way with your robot. I’d like to know wether you still use the compass on the new IMU and if so how you combine gyroscope, compass and accelerometer readings.
You mentioned that you are now using a 9-axis IMU. How much improvement did you get compared to the previous sensor? Can you give us a value? And how do you combine the measurements of all sensors to get the robot orientation?
Achieve eternal glory… Apenas. Só isso. Mais nada. Entendi.
Let the bar be raised!
Firstly, to get an absolute global direction, we need to calibrate the robot, lining up it to the earth’s magnetic north at a certain angle.
The gyroscope and accelerometer provide information about accelerations in all three directions and rotations around each axis. Gravity provides a background directly from the accelerometer, so we can do a pretty good job of tracking short term movements. However, to track the real position and orientation in space, the six-axis sensor is not quite there because of error accumulation on the measurements of each axis, and, over time, these errors can add up to drift in the absolute direction. We can deal with this problem by using one more absolute directional sensor – a 3-axis magnetometer. The extra magnetic field information allows the sensing algorithms to compensate for small shifts on the measurements over much longer periods of time, so we can track the absolute change in position and orientation more accurately.
Please, if you want to know more about,
it will be a pleasure to answer!
In our last participation on RoboCup Junior, our team had some problems in the goal position identification, because we only used the magnetometer. So, we decided to test some cameras with a 9-axis IMU to achieve better results.
About the Open MV, we chose this camera because it has a built-in microcontroller, this makes our job more efficient and quicker. And as we were testing, it proved to be efficient in the goal identification that we plan to use in the striker strategies so it can score by knowing the direction it must kick the ball.
Please, if you want to know more about,
it will be a pleasure to answer!
In our last competition, we had used a 3-axis magnetometer(HMC5883L), but we had a lot of problems to determine a global position orientation to find the goal. From past experiences, those problems were most associated with the influence of different factors on the sensors readings, which are very susceptible, for example: the magnetometer readings change drastically due to the influence of electromagnetic fields from different sources,example like the motors or being in contact with conductor materials (on its base to support it, for instance).
When we added the new 9-axis IMU(LSM9DS1) with the Open MV camera we didn’t had much time to test it because of the pandemic, but from the tests we held, we empirically concluded that we get better global robot orientation (with the readings not fluctuating as much as it did with the magnetometer only). When we come back from the isolation, we will do more tests, then I can give you more precise values.
We are still planning to combine the measurements of the gyroscope and accelerometer that provide information about accelerations in all three directions and rotations around each axis and, the extra magnetic field information allows the sensing algorithms to compensate for small shifts over much longer periods of time, so we can track the absolute change in position and orientation more accurate.
Please, if you want to know more about,
it will be a pleasure to answer!
About the IMU sensors measurements combination, I am still studying different methods that we can apply in order to do it, such as the use of filters. So I can teach the students about it. Do you recommend any filter in particular?
Oh and recently I came across a paper of yours, @Felipe, concerning the Junior Soccer Simulation and I found it remarkable, so I would like to salute you for it!
We have some videos and pics of the robots on the last LARC (Latin American Robotics Competition) with our most recent robots, but it doesn’t showcase some of the new features that we applied before the social isolation due to the pandemic.
Felipe, I recommend checking the Kalman filter (and its variations) to combine the IMU sensors. This is probably the most classic algorithm for sensor fusion.
Nice to hear that you like the paper about the Junior Soccer Simulation! Unfortunately, we still didn’t implement it in the official competitions. I guess the system was still too crude for that. Anyway, I still think we should do it, maybe using an off-the-shelf simulator like CoppeliaSim or Webots.
