Artificial Intelligence coursework, designing and developing an AI for the Robocode game. For more info, please refer to the project report.
Other than Robocode, which is required to run the AI simulation, the robot has no further dependencies than standard Java libraries.
With Robocode running, access the Source Editor using
CTRL+E
From here, locate Wilde.java and use the Editor's built-in compiler to compile and package the robot with
CTRL+B
Save the file and, when prompted, click YES to save the compiled robot within Robocode's own robots directory.
With Robocode running, begin a new battle using
CTRL+N
From the available robots, add u1409675.Wilde to the list of selected robots. Feel free to add any other robots to fight against, in melee (a multi-bot free-for-all battle) or 1-vs-1 combat.
Click Start Battle to then begin the simulation.
Wilde is a solution to the Robocode challenge which extends the Simple Robot class - this means it can only perform actions sequentially and is unable to target, move and fire concurrently. A detailed report is provided, covering the design and implementation of the robot's AI in detail.
As an overview, the AI is comprised of three main areas:
The targeting is Guess Factor-based, implementing a wave-based virtual bullet system to quickly learn possible bullet hit angles based on enemy movement patterns. It prioritises the closest known enemy as its target and is versatile across both melee & 1-vs-1 battles.
The AI is segmented to allow more accurate guesses in 1v1 combat, where scans are more frequent.
In melee battles, the movement operates a Minimum Risk algorithm, calculating which direction will be the least dangerous to move into based on the locations and states of its enemies.
In 1-vs-1 combat, a Psuedo-Random pattern is adopted, attempting to remain perpendicular where possible for maximum escape angle from incoming bullets while moving random distances to counter pattern recognition-based targeting.
The radar uses a Weak Target Lock, attempting to remain focused on its target but scanning the entire battlefield when said target is lost - this happens often - to either find the existing, or a better (closer) target.