Index New states Heading problem Fixed and finished New states Before, we treated the distance to the target as an absolute value, which made sense because with a holonomic architecture ...

Index Training Training For the complete parking manoeuvre, several actions have been added to allow acceleration in either direction and turning at the same time. With this new implementati...

Index Finding issue Modifying classes to enable going backwards Modifying training Finding issue Smarts does not explicitly explain how to use the different architectures for the vehicle...

Index Test model Test model Training plot: A program has been created that simulates the scenario 100 times and calculates an approximate average success rate. Each time the parking is su...

Index Plotting upgrades Desaligment class More testing Upgrading DQN Plotting upgrades The program has been modified to add two new graphs, the final horizontal and vertical distance, ...

Index Mountain car Working DQN Mountain car The most common neural network, with two hidden layers. class DQN(nn.Module): def __init__(self, state_size, action_size): super(DQN...

Index Fixing Reference System Working DQN Fixing Reference System Before we could do all this we had to fix the way of using the coordinates in SMARTS, as we were interested in a reference...

Index Finding the optimal parking spot Starting DQN Finding the optimal parking spot To make it more robust, the points are now divided into two sets with a minimum distance (the distance ...

Index Finding the optimal parking spot Modifying states Modifying rewards Coordinate and orientation problem Finding the optimal parking spot For now a simple solution has been created...

Index Decreased speed discretisation Fixed RLidar Example and data analysis Deeper explanation Decreased speed discretisation The agent accelerates by increasing its speed by 0.1. This...