This robot was constructed and raced in an annual team competition for ENMT221. We designed all aspects of the robot including the PCB. The race track consists of a black line on a white background. The robot must successfully follow the line all the way around, and the time taken determines the ranking.
In the weeks before the competition, we were quite busy with other projects, and we did not make much progress with the robot. We made a big push the night before the competition, fixing code, re-soldering connections and improving the mechanical design several times. By the time of the competition, we had managed to bring the robot to a fully functional state. The robot achieved second place overall, missing out on the first-place spot by only 0.8 seconds.
The design files are not currently public as this is a recurring course assignment.
The key features of the chassis are:
- Rail allowing adjustable placement of sensor boards. We found some arrangements caused the signal to be too discontinous.
- Batteries mounted underneath PCB to keep the center of gravity low.
- Adjustable motor mounts to vary the tension in the rubber bands used to turn the wheels.
- Silicone coating on the outside of the wheels for extra grip.
We found that a shorter chassis greatly improved the performance of the robot as it was able to make tight turns more easily.
The sensor boards each contain an IR led and sensor. The voltage in the sensing circuit varies with the amount of IR reflected. This analog circuit gave us a big advantage over other groups who could only read a digital "black" or "white" signal. With analog values, we can determine where the line is with greater accuracy as the sensor will return a value somewhere between the maximum and minimum.
The main board has the MCU (atmega4808) which reads the information from the sensor boards and decides what to do. The motors are driven with power mosfets.
I made an attempt at creating a mosfet gate driver circuit with an LM555 timer. Due to the competition rules, this was the only available chip I could use.
The circuit was intended to switch the mosfet faster with higher current - in theory this would make the system more efficient. The circuit worked but did not provide any noticeable increase in performance.
The motor driver circuit only allowed for the motors to turn in one direction. Designing an h-bridge would be more complex, requiring charge pumps or some other method to boost the voltage to the high-side mosfets. However, an h-bridge would allow the motors to spin backwards, which mean the robot can turn much tighter. In our case, this was not necessary to get a good score as no other group succesfully implemented an h bridge system.
The information from the sensors was used to determine how far the robot had deviated from the line. Because our circuit used analog values, this produced a relatively continous and accurate result. This error value is then fed into a PID controller which produces an output for the motors. This output is scaled then converted to a PWM signal for each motor, depending on the direction.
Last minute tuning of the PID controller was frustrating and seemed to produce variable results. This was likely due to variations in light causing different results on the IR sensors. In the future, a filter system could be implemented to reduce noise. Additionally, the performance of the motors varied with battery charge, so we could also implement a system to compensate for this.