Plane simulator for particle filter explanation

Description

Base of program allowing to create particles to localise the plane

TP réalisé par Edoaurd MALLECOURT et Jean-Baptiste PORRET

Requirement

Create Virtual Env (python3)

• Create a virtual env at the repository root

python -m venv venv

• Swtich to virtual env

source ./venv/bin/activate      

• To go back to host env simply execute command below:

deactivate

Install requirments

• When venv is created and activate, install python dependencies:

pip install -r requirements.txt

How to start

• Start the simulation

chmod +x ./scripts/Plane_Simulation.py
cd scripts
./Plane_Simulation.py

Simulator command

• Space bar: pause resume the simulator
• Left click: Add an obstacle to the environment
• Right click: Remove an obstacle to the environment
• S key: Save the current environment (obstacles) into '\tmp\obstacles.npy'
• R key: Reset Plane position and particles filter
• + key: Speed up the plane
• - key: Slow down the plane

Program Behavior

The job to do

First Particule Filter

Update the file Particle_Filter.py to:

1. Create first particles set

def getRandParticle(self,nbr, start_x, max_x, start_y, max_y):
        particle_list = []
        ###################################
        ##### TODO
        ##   nbr: number fo particles
        ##   start_x: min x possible coordinate
        ##   max_x: max x possible coordinate
        ##   start_y: min y possible coordinate
        ##   max_y: max y possible coordinate
        #####
        ## Use the Particle object to fill the list particle_list
        ##

        return particle_list

2. Weight the particles

def weightingParticle(self,p_x, p_y, observed_distance):
        ###################################
        ##### TODO
        ##   p_x: x coordinate of the particle p
        ##  p_y: y coordinate of the particle p
        ##  observed_distance: distance to the ground
        ##  measure by the probe
        ##
        ## return weight corresponding to the given particle
        ## according observation
        ##
        ## Note ue the function distance_to_obstacle to get the
        ## estimate particle to the ground distance
        return ""

3. Select particles according the weights

def weighted_random_choice(self,choices):
        ###################################
        ##### TODO
        ##   choices: dictionary holding particle coordination as key
        ##  and weight as value
        ##  return the selected particle key
        #####
        return ""

4. Resample particles

def motion_prediction(self):
        new_particle_list = []
        choices = {}
        for i in range(len(self.particle_list)):
            choices[self.particle_list[i].id()] = self.particle_list[i].w

            ###################################
            ##### TODO
            ##   self.particle_list: list of available particles
            ##
            #####
            ## Use the function self.weighted_random_choice(choices) returning
            #  coordinate from a particle according a
            ##  roulette wheel algorithm
            #  Note that weighted_random_choice return a string containing coodinate x and y of the selected particle
            #   coord = self.weighted_random_choice(choices)
            #   x_coord = int(coord.split('_')[0])
            #   y_coord = int(coord.split('_')[1])

        return new_particle_list

Influence of environement and different parameters

• Generate several test environments (at least 5) and highlight particle filter behaviour of each of them (specify the chosen, weight computation method, motion and observation error model)

• Test different weight computation method (at least 2) on these generated environments (specify the chosen motion and observation error model)

• Test different motion models (at least 2) error on these generated environments (specify the chosen weight computation method and observation error model)

• Test the influence of the observation error model on these generated environments (specify the chosen weight computation method and motion error model)

For each analysis, provide a set of screenshots and your associated comments and conclusion

Example of result (video)

Advanced Mode

Context

• Improve your particle filter by adding moves capacity to the plane.
• By clicking on the up arrow and down arrow, the plane goes up or go done
• Generated particles no more stay on the same y coordinate and need to be generated and move on both x and y coordinates

the job to do

• Modify the plane simulator in order to allow plane move on both x and y axes
• Modify particle filter to word on theses new conditions