This repository contains the code and general description for a robot designed to autonomously play "basketball". This work was done for the final project of ME348E/392Q - Advanced Mechatronics at The University of Texas at Austin during the Spring, 2022 semester. The project is adapted from Ed Carryer, instructor of ME210 at Stanford University. This design was able to win the final competition and set a new record for the most number of baskets made in a single game.
The final robot design for this project can be seen in the picture below. It is constructed primarily from 3D printed parts with lasercut wood also used in the construction of the shooter. The base, turret mechanism, and sensor and electronics mounting was designed by Elliott Turner and the shooter by Yoon Lee. Wiring was also done by Elliott Turner. The robot's operation can be broken down into 5 subsystems and follows a 2 stage strategy when operating.
The 5 subsystems handle the various functions of the robot and work together during competition to enable successful performance.
The ultrasonic subsystem is responsible for using the ultrasonic sensor to take distance measurements. The ultrasonic sensor is mounted stationary to the base of the robot and angled such that it measures the distance to the wall of the court. The sensor sends out a short burst of ultrasonic sound waves and measures the time taken for the echo to be detected. This time and the speed of sound are used to calculate the distance to the surface off of which the sound was reflected.
The built-in motors and encoders on the RSLK platform are used to move the robot. Independent PID loops were used for each motor to maintain motor speed across different surfaces (painted wood and electrical tape). The encoders were also used to track the total number of revolutions to accurately track distance traveled or degrees turned.
Three IR sensors were mounted stationary to the robot base and aimed such that each would be focused on a single IR beacon when the bot was in the shooting position. Each IR sensor is enclosed in a housing that limits it's field of view. The housing interiors were coated with a non-reflective paint to minimize reflections causing unwanted active readings.
The turret uses a gear drive driven by a stepper motor to aim the shooter. It is mounted statically to the robot base and holds the turret. There is no encoder, but code was written to minimize jerk in order to avoid skipping steps, allowing the stepper motor alone to be sufficient for accurately maintaining aim.
The shooter is attached to the top of the turret and contains all necessary components for storing, feeding, and shooting balls. The back of the shooter acts as a tall, slightly tilted magazine for carrying extra balls. A geared feeder motor uses a rubber wheel to move balls from the magazine to the shooter. The feeder motor includes an encoder which is used to track the distance balls have moved through the feeder system. An IR proximity sensor is also used to track balls in the feeder system to handle for any inconsistencies in the distance between balls. The shooter itself consists of a larger DC motor attached to a flywheel via a timing belt. Another encoder is attached to the flywheel's shaft to read the flywheel speed and ensure that balls are shot at the correct velocity. A hood over the flywheel defines the shooting angle and also induces back-spin on the balls to increase the effectiveness of backboard shots and ultimately increase the tolerance for making a basket.
The two-phase strategy is designed to serve as a reliable method for successfully identifying and shotting the correct baskets. The first phase is focused on navigating the robot to a pre-defined shooting position and the second is focused on identifying and shooting baskets.
In order to be able to shoot baskets, the robot must first reach the correct shooting position and orientation. This was chosen to the be the center of the court, facing straight at the center basket. It is also known that the robot will always start at a random orientation somewhere in the starting zone at the back of the court. The general idea is to first determine orientation, then reach the center line, and finally follow the center line to the shooting position.
The two key features of the starting zone taken advantage of is that the closest wall will always be the back wall and that the center line is the only line passing through it. With that in mind, the first step is to use the back wall to set the robot orientation. This is done by taking a set of distance measurements across a range of angles and determining at what angle the distance reaches a minimum. This angle is safely assumed to be the orientation at which the robot is facing directly at the back wall. Then, if the robot is not already on the center line, it simply uses a distance measurement parallel to the back wall to determine which side of the center line it is on and drive to it. Then the robot follows the center line to the first intersection, and than moves forward a set distance to the shooting position.
While in theory this should be sufficient for getting the robot in the correct position and orientation, it is not. The influence of various sources of error stack up to cause the robot to be far enough off of the desired position to potentially miss baskets. The first step in addressing this is to use the center line to straighted the robot such that it points straight. This is done by simply measuring the offset between the robot center and the center line at a known distance in front of and behind the shooting position. These two values in conjunction with the distances they were taken at can be approximated as making up the sides of a right triangle allowing the angle between the robot's orientation and centerline to be calculated. The robot then turns accordingly. Now that it is straight, the ultrasonic sensor is used again to adjust the robots distance from the front wall so that it is at the correct distance for making baskets.
Now that the robot is in the shooting position, it's time to make some baskets! To do this the robot sits and waits for an IR beacon to become active. The robot then sets the turret angle to point at the active basket and shoots with the necessary power to make it. It then marks that this was the last basket shot at and waits for a different beacon to be active. This process repeats indefinitely (which in practice is until the last ball is shot and someone hits the reset button and removes the robot from the court).
block diagram of the overall system
flowchart TD
launchpad(TI Launchpad)
rslk_motor(RSLK Motors)
rslk_encoder(RSLK Encoders)
motor_driver(Motor Driver)
stepper_driver(Stepper Driver)
shooter_motor(Shooter Motor)
feeder_motor(Feeder Motor)
shooter_encoder(Shooter Encoder)
feeder_encoder(Feeder Encoder)
ir_prox(IR Proximity Sensor)
ultra(Ultrasonic)
ir1(IR Beacon Sensor 1)
ir2(IR Beacon Sensor 2)
ir3(IR Beacon Sensor 3)
subgraph ultrasonic
ultra
end
ultra-->launchpad
subgraph ir
ir1
ir2
ir3
end
ir1-->launchpad
ir2-->launchpad
ir3-->launchpad
subgraph motion
rslk_motor
rslk_encoder
end
launchpad-->rslk_motor
rslk_encoder-->launchpad
rslk_encoder-.-rslk_motor
subgraph turret
stepper_driver
stepper_motor
end
launchpad-->stepper_driver
stepper_driver-->stepper_motor
subgraph shooter
motor_driver
shooter_motor
feeder_motor
shooter_encoder
feeder_encoder
ir_prox
end
launchpad-->motor_driver
motor_driver-->shooter_motor
motor_driver-->feeder_motor
shooter_motor-.-shooter_encoder
feeder_motor-.-feeder_encoder
shooter_encoder-->launchpad
feeder_encoder-->launchpad
ir_prox-->launchpad
stateDiagram-v2
s1001: read bumper 4
s1002: read bumper 5
s2001: turn 10 deg CW
s2002: measure distance
s2003: turn 90 deg CCW
s2004: turn 4 deg CW
s2005: measure distance
s2006: turn to angle of minimum measurement
s2007: check for line
s2008: turn 90 deg CW
s2009: measure distance
s2010: turn 180 deg
s2011: drive to line
s2012: turn 90 deg CCW
s2013: turn 90 deg CW
s2014: turn 180 deg
s2015: follow line to intersection
s2016: follow line for 12 in
s2017: straighten on line
s2018: measure distance
s2019: move forward/back to set distance to 31 in
s3001: read IR sensors
s3010: set last basket to 'L'
s3011: set turret angle to -28 deg
s3012: shoot with power 780
s3020: set last basket to 'M'
s3021: set turret angle to 0 deg
s3022: shoot with power 750
s3030: set last basket to 'R'
s3031: set turret angle to 28 deg
s3032: shoot with power 780
[*] --> startup
state startup {
[*] --> s1001
s1001 --> load: yes
load --> s1002
s1001 --> s1002 : no
s1002 --> s1001: no
s1002 --> [*]: yes
}
startup --> navigation
state navigation {
[*] --> s2001
s2001 --> s2002
s2002 --> s2001: less than 16 in
s2002 --> s2003
s2003 --> s2004
s2004 --> s2005
s2005 --> s2004: has turned less than 180 deg
s2005 --> s2006
s2006 --> s2007
s2007 --> s2008: not already on line
s2008 --> s2009
s2009 --> s2010
s2010 --> s2011: distance greater than 36 in
s2011 --> s2012: distance greater than 36 in
s2009 --> s2011: distance less than 36 in
s2011 --> s2012: distance less than 36 in
s2011 --> s2013: distance greater than 36 in
s2007 --> s2014: already on line
s2012 --> s2015
s2013 --> s2015
s2014 --> s2015
s2015 --> s2016
s2016 --> s2017
s2017 --> s2018
s2018 --> s2019
s2019 --> [*]
}
navigation --> shooting
state shooting {
[*] --> s3001
s3001 --> s3010: left IR active and last basket not 'L'
s3010 --> s3011
s3011 --> s3012
s3001 --> s3020: middle IR active and last basket not 'M'
s3020 --> s3021
s3021 --> s3022
s3001 --> s3030: right IR active and last basket not 'R'
s3030 --> s3031
s3031 --> s3032
s3012 --> s3001
s3022 --> s3001
s3032 --> s3001
}
Below is a (poorly hand-drawn) schematic of the robot. Note that connections between the launchpad and RSLK motors are not shown as all associated wiring is built into the RSLK main board.
