Formula Student Driverless Path Planning Algorithm. Colorblind centerline calculation algorithm developed by FaSTTUBe. It introduces a novel approach that uses neither Delaunay Triangulation nor RRT.
License: MIT License
IndexError in sorting stage
[global_planner-4] all_end_configurations, history = find_all_end_configurations(
[global_planner-4] File "/usr/local/lib/python3.10/dist-packages/fsd_path_planning/sorting_cones/trace_sorter/end_configurations.py", line 512, in find_all_end_configurations
[global_planner-4] ) = _impl_find_all_end_configurations(
[global_planner-4] File "/usr/local/lib/python3.10/dist-packages/fsd_path_planning/sorting_cones/trace_sorter/end_configurations.py", line 397, in _impl_find_all_end_configurations
[global_planner-4] current_attempt[position_in_stack] = next_idx
[global_planner-4] IndexError: index 1 is out of bounds for axis 0 with size 1Currently every step of the pipeline runs for every function call. However when no new cones have been added to the map and the picked started cones are the same as in the last calculation we can use the results from the previous calculation and only compute the fine path that starts from the car. This will potentially improve performance especially when the car moves slowly and the pose doesn't change often.
Currently many parameters of the algorithm are either hardcoded or only adjustable in deep parts of the codebase. All parameters that can meaningfully affect either the algorithm's output or performance need to be made available in the public API. Furthermore, two parameter presets need to be made available, one for when the algorithm runs in colorless mode and one when colors are always available and are guaranteed to be mostly correct.
This will be the last year this repo will be intensely maintained by me. In order to make sure that the algorithm can be used in the future a thorough documentation of every aspect of the pipeline needs to be created. This needs to happen until the end of September which is the time after which I no longer plan to be involved in FS anymore.
Currently Skidpad relocalization process is obfuscated. However it is important for many Pipelines to have information about the relocalization. The PathPlanner public API needs to be updated so that the information is made available.
The information that need to be made publicly available are:
To encode the relocalization information a dataclass will be created that will contain that info.
The PathPlanner class will get a new method (probably property) that will return an instance of the new dataclass.
The issue will be updated if any amendments are needed.
I am using the fsd_path_planning package for a Skidpad mission and encountering a ZeroDivisionError during path calculation. Additionally, the path planner seems to ignore orange cones.
When executing the path planning for a Skidpad mission, a ZeroDivisionError occurs in the connect_path_to_car method. The error trace is as follows:
Traceback (most recent call last):
...
File "[...]/lib/python3.10/site-packages/fsd_path_planning/calculate_path/core_calculate_path.py", line 451, in connect_path_to_car
angle_to_first_point = vec_angle_between(...)
ZeroDivisionError: division by zero
I also noticed that the path planner does not consider the orange cones in its calculations. I am not sure if this is an intended behavior or a limitation in the current Skidpad implementation.
Python Version: 3.10
fsd_path_planning version: 0.3.2
I am attaching the relevant code snippet for reference. Any insights or fixes for this issue would be greatly appreciated.
import numpy as np
from fsd_path_planning import PathPlanner, MissionTypes, ConeTypes
import matplotlib.pyplot as plt
# Skidpad track
cones_blue = np.array(
[[0.0, 15.0], [-0.5804185646014384, 17.91796117178381], [-2.233310793452575, 20.391689206547426],
[-4.70703882821619, 22.044581435398563], [-7.624999999999999, 22.625], [-10.542961171783809, 22.044581435398563],
[-13.016689206547424, 20.391689206547426], [-14.669581435398563, 17.91796117178381], [-15.25, 15.000000000000002],
[-14.669581435398563, 12.082038828216191], [-13.016689206547426, 9.608310793452576],
[-10.542961171783814, 7.95541856460144], [-7.625000000000002, 7.375], [-4.707038828216189, 7.955418564601439],
[-2.233310793452577, 9.608310793452574], [-0.5804185646014401, 12.082038828216186],
[6.5589885311209155, 5.183779967067581], [10.624999999999998, 4.375], [14.691011468879081, 5.183779967067579],
[18.138009550107064, 7.4869904498929305], [20.44122003293242, 10.933988531120914], [21.25, 15.0],
[20.44122003293242, 19.06601146887908], [18.138009550107068, 22.513009550107068],
[14.69101146887908, 24.81622003293242], [10.625, 25.625], [6.558988531120922, 24.81622003293242],
[4.722066274166729, 23.834364630714546], [4.7220662741667265, 6.165635369285457]])
cones_yellow = np.array(
[[18.25, 15.0], [17.669581435398563, 17.91796117178381], [16.016689206547426, 20.391689206547426],
[13.54296117178381, 22.044581435398563], [10.625, 22.625], [7.7070388282161915, 22.044581435398563],
[5.233310793452576, 20.391689206547426], [3.5804185646014384, 17.91796117178381], [3.0, 15.000000000000002],
[3.5804185646014375, 12.082038828216191], [5.233310793452574, 9.608310793452576],
[7.707038828216186, 7.95541856460144], [10.624999999999998, 7.375], [13.542961171783812, 7.955418564601439],
[16.016689206547422, 9.608310793452574], [17.66958143539856, 12.082038828216186],
[-3.558988531120921, 24.81622003293242], [-7.624999999999999, 25.625], [-11.69101146887908, 24.81622003293242],
[-15.138009550107068, 22.513009550107068], [-17.44122003293242, 19.06601146887908], [-18.25, 15.000000000000002],
[-17.44122003293242, 10.933988531120923], [-15.13800955010707, 7.486990449892933],
[-11.691011468879084, 5.183779967067581], [-7.625000000000002, 4.375], [-3.558988531120919, 5.183779967067579],
[-1.7220662741667256, 23.834364630714543], [-1.72206627416673, 6.165635369285454]])
cones_orange_big = np.array([[0.0, 14.5], [0.0, 15.5], [3.0, 14.5], [3.0, 15.5]])
cones_orange_small = np.array(
[[0.0, 0.0], [3.0, 0.0], [0.0, 5.0], [3.0, 5.0], [0.0, 40.0], [1.0, 40.0], [2.0, 40.0], [3.0, 40.0], [0.0, 37.0],
[0.0, 34.0], [0.0, 31.0], [0.0, 28.0], [0.0, 25.0], [3.0, 37.0], [3.0, 34.0], [3.0, 31.0], [3.0, 28.0],
[3.0, 25.0]])
#create empty array for unknown cones
cones_unknown_raw = np.empty((0, 2), float)
def load_data():
global_cones = [cones_unknown_raw, cones_yellow, cones_blue, cones_orange_small, cones_orange_big]
car_position = np.array([1.5, 2])
car_direction = np.array([0, 1.0])
return global_cones, car_position, car_direction
# Load data and create path planner
global_cones, car_position, car_direction = load_data()
path_planner = PathPlanner(MissionTypes.skidpad)
# Calculate path
out = path_planner.calculate_path_in_global_frame(global_cones, car_position, car_direction, return_intermediate_results = True)
(
path,
sorted_left,
sorted_right,
left_cones_with_virtual,
right_cones_with_virtual,
left_to_right_match,
right_to_left_match,
) = out
# Plot results
plt.plot(*left_cones_with_virtual.T, "o-", c="b")
plt.plot(*right_cones_with_virtual.T, "o-", c="r")
plt.plot(*sorted_left.T, "o-", c="b")
plt.plot(*sorted_right.T, "o-", c="r")
plt.plot(*path[:, 1:3].T)
plt.axis("equal")
plt.show()
Currently only the Skidpad mission has relocalization abilities. We want to extend this to Acceleration and AutoX/Trackdrive.
For Acceleration we want to estimate the track width and the angle of the car relative to the Acceleration track direction.
For AX/TD we want to be given cone observation from a previous run and align it with the cones seen in this run.
Latest numba version seems to not be compatible/have a bug. Using numba <0.60 seems to work.
Create a system that given a set of tracks, applies an automated parameter estimation algorithm that finds parameters that maximizes the number of track configurations for which the algorithm works correctly. This way the correct parameters can be found for the track configuration one cares for.
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
Django
The Web framework for perfectionists with deadlines.
Laravel
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft
Open source projects and samples from Microsoft.
Google
Google ❤️ Open Source for everyone.
Alibaba
Alibaba Open Source for everyone
D3
Data-Driven Documents codes.
Tencent
China tencent open source team.