Difference between revisions of "S22: TBD"

Latest revision as of 06:43, 28 May 2022

Project Title

THE BEST DRIVER

Abstract

Our goal for this project is to use knowledge we gathered from lectures to design, implement, and test a self-driving RC car using a Controller Area Network (CAN) bus for controller communication. The project involves FreeRTOS and utilizes periodic tasks (running at 1Hz, 10Hz, and 100Hz) to gather, process, and display data from various embedded modules.

Introduction

The project was divided into 5 modules:

• Sensor/Bridge Controller
• Motor Controller
• Geo Controller
• Driver/LCD Controller
• Web Application

Team Members & Responsibilities

From Left to Right: Jasdip Sekhon, Brian Ho, Justin Stokes, Billy Lai, Isaac Wahhab, William Hernandez

• Sensor
  • Jasdip Sekhon
  • Brian Ho
  • Justin Stokes

• Motor
  • Justin Stokes
  • Isaac Wahhab

• Geographical
  • Brian Ho
  • William Hernandez
  • Jasdip Sekhon

• Driver Logic & LCD & LED Ring
  • William Hernandez
  • Billy Lai
  • Isaac Wahhab

• Web Application & Communication Bridge Controller
  • Isaac Wahhab
  • Billy Lai

• Testing Team
  • The Best Driver Team

Schedule

Parts List & Cost

Printed Circuit Board

GitLab link to PCB board (.brd) files and schematic files (.sch) created in EAGLE PCB design software

We designed and implemented two PCBs: CAN Bus PCB and RC Car PCB.

CAN Bus PCB Schematic

CAN Bus PCB Board

RC Car PCB Schematic

RC Car PCB Board

CAN Communication

Our message IDs are arranged with priority given to the nodes in the following order: DRIVER, SENSOR, GEO, MOTOR. The following critical messages are transmitted at 100Hz:

• SENSOR_TO_DRIVER_SONARS
• GPS_DESTINATION_LOCATION
• SENSOR_TO_DRIVER_WHEEL_ENCODER_MSG
• GEO_STATUS
• DRIVER_TO_MOTOR_CMD

The remaining debug messages are transmitted at 10Hz or 1Hz.

We decided to give 32 message IDs to each node and to separate the nodes' message IDs accordingly.

Hardware Design

DBC File

Gitlab link to our master branch's DBC file

Our DBC file includes the SENSOR, DRIVER, MOTOR, and GEO nodes.

Below are the messages defined in the file:

BO_ 32 DRIVER_TO_MOTOR_CMD: 3 DRIVER
 SG_ DRIVER_TO_MOTOR_steer : 0|9@1+ (1,-180) [-180|179] "" MOTOR
 SG_ DRIVER_TO_MOTOR_speed : 9|5@1+ (1,0) [0|31] "RPM" MOTOR
 SG_ DRIVER_TO_MOTOR_current_rpm : 14|10@1+ (1,0) [0|1023] "RPM" MOTOR

BO_ 33 DRIVER_DEBUG_MSG: 1 DRIVER
 SG_ DRIVER_DEBUG_MSG_sensor_mia : 0|1@1+ (1,0) [0|0] "" DBG
 SG_ DRIVER_DEBUG_MSG_geo_mia : 1|1@1+ (1,0) [0|0] "" DBG

BO_ 64 SENSOR_TO_DRIVER_SONARS: 4 SENSOR
 SG_ SENSOR_TO_DRIVER_SONARS_front_left : 0|8@1+ (1,0) [0|0] "" DRIVER
 SG_ SENSOR_TO_DRIVER_SONARS_front_middle : 8|8@1+ (1,0) [0|0] "" DRIVER
 SG_ SENSOR_TO_DRIVER_SONARS_front_right : 16|8@1+ (1,0) [0|0] "" DRIVER
 SG_ SENSOR_TO_DRIVER_SONARS_back : 24|8@1+ (1,0) [0|0] "" DRIVER

BO_ 65 GPS_DESTINATION_LOCATION: 7 SENSOR
 SG_ GPS_DESTINATION_LOCATION_latitude : 0|28@1+ (0.000001,-134.217727) [-134.217727|134.217728] "" GEO,DRIVER
 SG_ GPS_DESTINATION_LOCATION_longitude : 28|28@1+ (0.000001,-134.217727) [-134.217728|134.217728] "" GEO,DRIVER 

BO_ 66 SENSOR_TO_DRIVER_WHEEL_ENCODER_MSG: 2 SENSOR
 SG_ SENSOR_TO_DRIVER_WHEEL_ENCODER_MSG_rpm : 0|10@1+ (1,0) [0|0] "RPM" DRIVER,DBG

BO_ 67 SENSOR_TO_DRIVER_APP_CONNECTION_MSG: 1 SENSOR
 SG_ SENSOR_TO_DRIVER_APP_CONNECTION_MSG_app_connected : 0|1@1+ (1,0) [0|0] "" DRIVER 

BO_ 96 GEO_STATUS: 5 GEO
  SG_ GEO_STATUS_compass_heading : 0|9@1+ (1,0) [0|359] "Degrees" SENSOR,DRIVER
  SG_ GEO_STATUS_destination_heading : 9|9@1+ (1,0) [0|359] "Degrees" SENSOR,DRIVER
  SG_ GEO_STATUS_distance_to_destination : 18|16@1+ (0.1,0) [0|0] "Meters" SENSOR,DRIVER

BO_ 97 GEO_DEBUG_MSG: 1 GEO
  SG_ GEO_DEBUG_MSG_lock_status : 0|1@1+ (1,0) [0|0] "" DRIVER,DBG
  SG_ GEO_DEBUG_MSG_num_satellites : 1|4@1+ (1,0) [0|15] "" DRIVER,DBG

BO_ 98 GEO_CURRENT_LOCATION_DEBUG: 7 GEO
  SG_ GEO_CURRENT_LOCATION_DEBUG_latitude : 0|28@1+ (0.000001,-134.217727) [-134.217727|134.217728] "" DRIVER,DBG
  SG_ GEO_CURRENT_LOCATION_DEBUG_longitude : 28|28@1+ (0.000001,-134.217727) [-134.217728|134.217728] "" DRIVER,DBG

BO_ 99 GEO_CURRENT_DESTINATION_LOCATION_DEBUG: 7 GEO
  SG_ GEO_CURRENT_DESTINATION_LOCATION_DEBUG_latitude : 0|28@1+ (0.000001,-134.217727) [-134.217727|134.217728] "" SENSOR,DBG
  SG_ GEO_CURRENT_DESTINATION_LOCATION_DEBUG_longitude : 28|28@1+ (0.000001,-134.217727) [-134.217728|134.217728] "" SENSOR,DBG

BO_ 128 MOTOR_DEBUG_MSG: 3 MOTOR
 SG_ MOTOR_DEBUG_MSG_echo_steer : 0|3@1+ (1,-2) [-2|2] "" DBG
 SG_ MOTOR_DEBUG_MSG_echo_speed : 3|5@1+ (1,0) [0|31] "RPM" DBG
 SG_ MOTOR_DEBUG_MSG_echo_rpm : 8|10@1+ (1,0) [0|1023] "RPM" DBG

MIA Management

Since the DRIVER node is our main controller, we conducted MIA management in the driver_logic code module. Our MIA management is serviced at 10Hz and ensures that the expected messages are received within a time limit set by threshold values. If a message goes "missing" past set threshold times, a default replacement value is used.
The MIA threshold values and replacement structures are found in the can_mia_configurations.c file.

SENSOR_TO_DRIVER_SONARS

• Default value in the DRIVER node is to set the sensor values to 0. This will result in the driver_logic simulating obstacles detected at 0cm in the front and back, and the car will not move.

GPS_DESTINATION_LOCATION

• Default value in the GEO node is to set the current coordinates and destination coordinates to 0. This will result in a 0 distance_to_destination calculation, and the DRIVER node will subsequently send an IDLE speed to the MOTOR node.

GEO_STATUS

• Default value in the DRIVER node is to set compass heading, destination heading, and distance to destination to 0. This will result in the ring LED not lighting up (to show an invalid state) and the DRIVER will send an IDLE speed to the MOTOR node.

DRIVER_TO_MOTOR_CMD

• Default value in the MOTOR node is to set the motor speed and steering to IDLE and 0, respectively. Without updated driver to motor commands, the RC car should not move.

Sensor Controller

Sensor Node GitLab

Hardware Design

We used 4 Gravity DFRobot URM09 ultrasonic sensors. These ultrasonic sensors are used by the RC car for the purpose of obstacle avoidance. Three sensors were placed on the front side of the RC car and one on the back. They use I2C communication to send distance data to the SJ2 microcontroller. On the SJ2 microcontroller, P0.10 was used for SDA, and P0.11 was used for SCL. The measurement range is 300cm. It takes in 3.3V supply voltage. Using the I2C driver, the registers on the ultrasonic sensors were configured and the distance was read from the specified registers in the datasheet.

Wheel Encoder

The wheel encoder is interfaced using 3 pins: VCC, GND, and OUT. The voltage input requires 4.5-5.5V and the data pin outputs 5V. The 5V output is converted to 3.3V using a level converter, and then used as a GPIO input to P0.22 on the SENSOR node.

The rotary wheel has 20 slits spaced evenly. An infrared beam and sensor is used to detect when the beam is blocked. When the beam is blocked, the data OUT pin outputs 0V. When the beam is unblocked, the data OUT pin outputs 5V.

Software Design

The ultrasonic sensors use the I2C driver available in the project. The implementation is done in the 100Hz periodic callback function. After writing to the control register address of the ultrasonic sensors, the distance values are read from another register which provided the distance in a single byte. To prevent signal interference, only one sensor sends pulses for each iteration of the periodic_callback. After the distance values were read, they were sent to the driver for obstacle avoidance. The following are the functions used to configure the ultrasonic sensor by writing to the register and then reading from the register specified in the sensor datasheet over I2C:

static const uint8_t distance_low_bits_register = 0x04;
void ultrasonic__passive_measurement_mode(uint8_t slave_address) {
  const uint8_t passive_measurement_mode = 0x00;
  const uint8_t control_register = 0x07;
  i2c__write_single(I2C__2, slave_address, control_register, passive_measurement_mode);
}

uint8_t ultrasonic__single_sensor_read(uint8_t slave_address) {
  return i2c__read_single(I2C__2, slave_address, distance_low_bits_register);
}

Wheel Encoder

The wheel encoder uses a GPIO pin (P0.22). To measure the speed of rotation, we enabled GPIO interrupts by adding a gpio_isr module to dispatch GPIO interrupts to associated handler methods. On rising edges, a GPIO interrupt service routine (ISR) increments a wheel encoder rising edge counter. In the 100Hz periodic callback, the wheel encoder counter is used to calculate the measured rotations per minute (RPM). This RPM value is sent from the SENSOR node and is relayed through the DRIVER node to the MOTOR node. After sending out the RPM value, the wheel encoder rising edge counter is reset.

Technical Challenges

1. Ultrasonic sensor interference

• Since we had 3 ultrasonic sensors on the front side sending sound waves, there was interference between them. As a result, the values being read from the sensors were incorrect. The solution to this was to only read one sensor per periodic callback. In doing so we were able to obtain accurate sensor readings, but they were slower than what we needed for our obstacle avoidance logic. to counteract the slower measurements, we ended up moving the signal readings to the 100Hz callback instead, still reading one sensor per callback.

2. Ultrasonic sensor wiring not secure

• One of the issues we faced was that the sensors would not be connected properly because the RC car was constantly being moved around. A solution was to clean up the wiring and use a PCB. Also, we would use CAN Busmaster to check which sensors were not connected.

3. Wheel Encoder Debouncing

• When we hooked up the wheel encoder we didn't think about debouncing. Since the values were coming in fast enough, the motor logic hid the fact that sometimes we were getting spikes in the rising edge counter which made moving the car at slower speeds much more difficult. If we had noticed this sooner, we would have used a hardware solution and create a deboucing circuit on the PCB. Since we didn't have time to design and order another PCB, we went with a software debouce instead.

Motor Controller

Motor Controller GitLab

Hardware Design

The hardware design for the motor controller was determined by the hardware of the car. We knew that the cars motors were controller through an electronic speed controller(ESC) and just had to interface the SJ2 with this. The top of the receiver revealed 12 pins, in 4 rows of 3. Each row had the following according the the sticker: GND, VCC, and signal. The signal pin would be hooked up to the pins which are firmware configured to output PWM.

Software Design

Controlling the motors was done through firmware generating PWM signals. We decided to begin by hooking up the car's receiver to a Saleae Logic Analyzer to see the waveforms generated when using the remote control that came with the car. This was helpful in many ways. First, we were able to see which channel was used for the steering servo, and which was used for the actual motor. Second, we were able to see the maximum, minimum, and neutral PWM duty cycles required and the period period of the waveform. Lastly, we were able to visually see the different sequences necessary to run the motors.

Steering Algorithm

A simple algorithm was used to determine the PWM signal sent to the steering servos. We decide to have the driver node send a direction between the values of -180 to 179. After some quick testing with the values, we found the range to be much larger than then we required to get the steering to be a fluid motion. To solve this, we had our firmware treat any signal above 40 and below -40 to be max right and left respectively. We then had to scale the values within that range using the algorithm in the flow chart above.

Steering Algorithm code snippet showing the scaling duty cycle outside of the +/- 40 range sent from driver.

void motor_controller__set_direction(int16_t direction) {
  if (direction < -40) {
    motor_controller__set_direction_pwm_cycle(11.96); // 6.66
  } else if (direction > 40) {
    motor_controller__set_direction_pwm_cycle(6.66);
  } else {
    float duty_cycle = (9.31 - (direction / 40.0) * 2.65);
    motor_controller__set_direction_pwm_cycle(duty_cycle);
  }
}

Forward Backward Algorithm

The motor speed control algorithm was more complicated due to the ESC having built in firmware to prevent the user from repeated sending forward and reverse signals and frying the motor. A specific routine was needed to allow the motor to correctly receive a reverse signal. This routine is depicted in the flow chart below. Other than that we used a fairly straight forward algorithm based on the neutral, max forward, and max reverse duty cycles read from the Saleae analyzer mentioned earlier. The driver would send a drive_speed from 0 - 31 and a current_rpm based on sensor readings. The 0 - 31 values were broken down as such:

• 0 and 5 were mapped to 'neutral' or 'idle' pwm cycles.
• 1 - 4 were varying speeds of reverse with 1 being fastest, and 4 being slowest reverse speed.
• 6 - 31 were varying speeds of forward with 6 being slowest, 31 being fastest.

The current_rpm was used to decided whether or not the car was moving at an expected speed based on a mapping which we allocated statically. From there the algorithm designed was pretty straight forward. Based on the speed and current rpm received from driver, we followed the above flow chart incrementing or decrementing the previously set PWM duty cycle as necessary. This algorithm could definitely have been improved using some basic control theory such as a PID controller, but we found that we were able to get a consistent and smooth ride speed without one.

Lookup table containing the targetted RPM for each speed sent by the driver.

static uint16_t rpm_table[] = {
    [0] = 0,    [1] = 300,  [2] = 250,  [3] = 200,  [4] = 175,  [5] = 0,    [6] = 175,  [7] = 200,
    [8] = 225,  [9] = 250,  [10] = 275, [11] = 300, [12] = 325, [13] = 350, [14] = 375, [15] = 400,
    [16] = 425, [17] = 450, [18] = 475, [19] = 500, [20] = 525, [21] = 550, [22] = 575, [23] = 600,
    [24] = 625, [25] = 650, [26] = 675, [27] = 700, [28] = 725, [29] = 750, [30] = 775, [31] = 800,
};

Flow chart explaining the reverse routine followed to enter a reverse state.

Technical Challenges

• Initial run through of the algorithm to control speed ended up 'sling-shotting' the car. This due to the PWM duty cycles being set would over shoot the desired RPM values by a large margin. To solve this, we changed the the value we scale up or down the PWM to 0.005f. Since our motor commands were being processed so quickly, a small value was needed to prevent the car from the 'sling-shotting' behavior of ramping up and down the motor to quickly.
• The reverse routine and start up routines of the motor may vary depending on the firmware that comes along with the ESC. For Traxxas, a specific routine was needed to start and reverse the motors. Without a logic analyzer or oscilloscope to visualize the routine, deciphering it would have been near impossible.
• Start off slow and then ramp up after. It really does not take much higher than the neutral duty cycle to make the cars move at a decent pace. Early on, we capped our speed to 300rpm and this would have been plenty fast enough to even get through the demo.

Geographical Controller

Code for GEO Node
Code for LSM303 Compass

GPS module

Compass and Accelerometer

Hardware Design

SparkFun Venus GPS with SMA Connector

For the GPS location, we used the SparkFun Venus GPS. The GPS module was interfaced using UART protocol with a baud rate of 9600 bps. The module was interfaced with an NMEA command to output GPGGA messages to the autonomous car.

LSM303DLHC Magnetometer

For the compass heading we used the lsm303DLHC compass + accelerometer from Adafruit. The compass was interfaced using I2C with a speed of 100KHz. The data output rate was set to 220HZ. The gain was left as the default +-1.3 gauss. The operation mode was switch to continuous conversion mode. The compass was placed at a safe distance above all of the other hardware components to minimize the magnetic disturbance.

Software Design

GPS Coordinate Message Parsing

The GPS module will output several UART messages containing geological location information. We set the message type to GPGGA format using the NMEA command provided by the documentation of the GPS module. The message string was parsed using the "strtok" function and hashed using the "atof" function to turn each part of the string into float values that the care can use.

Setting the NMEA command was done using the following function:

void gps_config__init_nmea_gpgga(void) {
  uint8_t message_data[100] = {0xA0, 0xA1, 0};
  uint8_t cs = 8;
  uint16_t length = 9;
  message_data[5] = 1;
  message_data[6] = 0;
  message_data[7] = 0;
  message_data[8] = 0;
  message_data[9] = 0;
  message_data[10] = 0;
  message_data[11] = 0;
  message_data[12] = 0;
  for (int index = 5; index < length + 4; index++) {
    cs = cs ^ message_data[index];
  }
  message_data[2] = length / 256;
  message_data[3] = length % 256;
  message_data[4] = 8;
  message_data[4 + length] = cs;
  message_data[5 + length] = 0x0D;
  message_data[6 + length] = 0x0A;

  for (int i = 0; i < 7 + length; i++) {
    uart__polled_put(UART__1, message_data[i]);
  }
}

An example of parsing and hashing the string:

static char *token;
const char *delim = ",";

token = strtok(gps_final_string, delim);
// get time
token = strtok(NULL, delim);
if (token != NULL)
  time = atof(token);

A GPGGA message appears as follows:

Message Format:
$GPGGA,HHMMSS.SS,DDMM.MMMMM,K,DDDMM.MMMMM,L,N,QQ,PP.P,AAAA.AA,M,±XX.XX,M, SSS,RRRR*CC<CR><LF>