This takes some direct inspiration from PythonRobotics (https://github.com/AtsushiSakai/PythonRobotics) and their example DWA control and markov localization.
To run the simulation, run main2.py.
There are multiple configurable parameters. The number of bots is configurable at the top of main.py.
Simulation configuration is available in the RobotConfig class at the top of robot.py. To switch the point generators,
you can change the class used during initialization of all robots -- for example, change PointFilter to GPSLocalizer.
You can change things marked with CONFIG.
PyRobo splits the movement pipeline into several abstract classes.
MotionModel: How to represent a robot given input. Currently we support YawMotionModel, which simulates a robot with significant inertia and slow-to-change yaw and speed.
PointGenerator: How to get predicted robot states. Currently we have GPSLocater which generates gaussian predictions around a robot's true position, and ParticleFilter, which also uses distances to known landmarks.
Localizer: How to generate a 'footprint' for the robot based on points generated. Currently, we support COCALU as described here: https://link.springer.com/article/10.1007/s10514-018-9726-5
Cost: A cost to impose on potential robot trajectories. Multiple are supported.
Planner: How to choose and evaluate paths. Currently we use the Dynamic Window Approach (DWA) based on predicted robot state, choosing the lowest-cost trajectory. The planner also supports a state machine, such that when costs meet thresholds a new robot state is triggered with different weights. The example code has three states: "nospace", "default", and "goal".
This simulation requires matplotlib, numpy, and scipy.