This repository contains all the code files for ESP8266 microcontroller regarding the BTW-Robot project.
The other part of code is the android app to control the robot remotely. LINK TO REPO
The main working principle of a self-balancing robot is to continuously adjust its movement in order to maintain a specific position and/or velocity.
The BTW-Robot utilizes two stepper motors which are driven to counteract its inertial movement.
The most important information for such a robot is the current angular position. To get angle data I used MPU6050 inertial measurement unit. The sensor consists of both gyroscope and accelerometer (both 3 Axis).
Accelerometer measures acceleration and since under relatively static conditions the only significant acceleration is gravitational, we can use that measurement to estimate the tilt angle of the robot.
Although the accelerometer would give us a pretty decent estimation of angle in static conditions, the method fails if any oscillations occur.
Gyro measures angular velocity and it does so quite precisely. Precisely enough to be the only source of inertial measurements for our purpouse. However there are two reasons we cannot use the gyro data alone:
- angular velocity is the derivative of angle and derivatives give no information about starting position. At least once we need to know actual angle, then we could rely on velocity
- gyro measurements tend to drift due to various long term noise (like Coriolis force), so even when no movement the sensor would show changing angle value over time
An easy, yet very effective solution is to combine accelerometer and gyro data. Accelerometer doesn't like high frequecies so we'll give it a low-pass filter. Gyro doesn't like low frequencies so we'll implement a high-pass filter.
angle = c * (angle + gyro * dt) + (1 - c) * accelerometer
where c would be around 0.98 (depends on time constant).
The above solution is called complementary filter.
We've got two closed-loop PID regulators working together in a cascade. The input for the outer regulator is speed error. Error is obtained by subtracting target value and estimated velocity value. The latter involves both absolute motor speed and relative head's speed. This way we can effectively take into account robot tilting as well.
I used WiFi feature of ESP8266 to remotely control the robot. Android app sends JSON packets containing steering data via UDP protocol to the ESP. In return ESP sends packets containing speed and tilt values the same way.