Skip to content

Grows trees including the features of tying and pruning based on pre-defined rules. Generates 3D models of these trees

Notifications You must be signed in to change notification settings

OSUrobotics/lpy_treesim

Repository files navigation

TreeSim_Lpy

Description

TreeSim_Lpy is a tree modelling tool which is built upon L-py with the added features of pruning and tying trees down to mimic different architectures. The tool uses python and prior knowledge of L-systems and L-Py is needed to work with this tool. This tool is ideal for researchers and developers working on botanical simulations and robotic harvesting applications.

Python version 3.9

Table of Contents

Documentation

The documentation is provided at Read the Docs.

You can find the latest L-Py documentation at <https://lpy.readthedocs.io/en/latest>

Installation

To install TreeSim_Lpy, follow these steps (adapted from the L-Py documentation):

  1. Install Conda:

    The L-Py distribution relies on the conda software environment management system. If you do not already have conda installed, you can find installation instructions on the Conda Installation Page.

  2. Create a Conda Environment:

    Create an environment named lpy:

    conda create -n lpy python=3.9.13 openalea.lpy -c fredboudon -c conda-forge

    The package is retrieved from the fredboudon channel (development), and its dependencies will be taken from the conda-forge channel.

  3. Activate the L-Py Environment:

    conda activate lpy
  4. Install Required Packages:

    pip install -r requirements.txt
  5. Run L-Py:

    lpy

Tutorials

There are many things you may want to modify as you grow your own trees. Here are some tutorials for some of the more common changes:

  1. Changing Apple Geometry:

    The call production of the apples happens in the grow_object(o) section:

    elif 'Apple' in o.name:
        produce [S(.1/2, .09/15)f(.1)&(180)A(.1, .09)]

    The apple's base is generated with the A(bh, r) production rule seen below.

    A(bh, r):
        curves = make_apple_curve()
        base_curve = curves[0]
        top_curve = curves[1]
        nproduce SetColor(230,0,0) SectionResolution(60)
        produce nF(bh, .01, r, base_curve) ^(180) nF(bh/5, .1, r, top_curve)^(180)

    The parameters represent the base height of the apple and the radius of the apple. If you wanted to create a completely new apple geometry, just replace the code in this A section. However, if you simply want to edit the existing shape of the apple, that can be done in the make_apple_curve() section.

    The apple is made with two curves: a curve that marks the base of the apple, and a curve that marks the indentation on top of the apple. These curves are generated as different Curve2D objects, then turned into QuantisedFunction objects. This is necessary because of the way the apple is produced nF. nF has an optional parameter radiusvariation which must be a quantized function. nF produces a cylinder in n steps, and these curves work by specifying how large the radius for the cylinder should be at each step.

    Currently, the stem is produced separately from the apple base. The stem is created in a slightly different way than the apple. A NurbsCurve2D object is returned from the make_stem_curve() function. This curve is used in SetGuide to mark how the stem will be generated. nF is used to follow the guide while generating a cylinder, and there is no radiusvariation this time.

    S(sh,r):
        stem_curve = make_stem_curve()
        nproduce SetColor(100,65,23)
        produce  SetGuide(stem_curve, sh) _(r)nF(sh, .1, r)
  2. Changing Leaf Geometry:

    The call production of the leaves happens in the grow_object(o) section:

    elif 'Leaf' in o.name:
        produce L(.1)

    Here, .1 is just a hard-coded value that doesn't have much significance.

    The actual generation of the leaf can be seen in the L(l) production section:

    L(l):
        nproduce SetColor(0,225,0)
    
        curves = make_leaf_guide()
        curve1 = curves[0]
        curve2 = curves[1]
        produce _(.0025) F(l/10){[SetGuide(curve1, l) _(.001).nF(l, .01)][SetGuide(curve2, l)_(.001).nF(l, .01)]}

    The parameter here serves as the length of the leaf. To edit the shape of the leaf, edits can be made in the make_leaf_guide() function. In the make_leaf_guide() function, a BezierCurve2D and its inverse are generated. These are returned to be used as guides for the SetGuide function provided by L-Py. These curves are generated with random points from a set range in order to create leaves of different shape. These randomness of these points, or the range they are generated from, could be edited to change the leaf shape. However, a new geometry altogether could be made in the L(l) section.

  3. Changing Bud Geometry:

    The call for the production of the buds occurs in the grow_one(o) section by addicting spiked_bud(o.thickness) to the produce call:

    if 'Spur' in o.name:
        produce I(o.growth_length, o.thickness, o) bud(ParameterSet(type=o, num_buds=0)) spiked_bud(o.thickness)grow_object(o)

    The actual bud is produced in the spiked_bud production section:

    spiked_bud(r):
        base_height = r * 2
        top_height = r * 2
        num_sect = 20
        produce @g(Cylinder(r,base_height,1,num_sect))f(base_height)@g(Cone(r,top_height,0,num_sect))

    This is one of the most basic objects generated on the tree model. As the buds on actual trees are little spikes, the bud geometry is made up of a cone on top of a cylinder. These are both produced with L-Py's basic @g primitive used to draw PglShapes. The height of the two shapes scale with the radius (provided as the parameter). The num_sect is used to determine how many sections each shape is made up of, and 20 was chosen as they appear circular without adding too many triangles to the model.

  4. Changing Branch Profile Curve:

    Every branch on the model has a unique profile curve. This is so the model doesn't appear to be perfectly cylindrical (as if made of PVC pipes). Every branch has its own unique curve that is generated when the branch is originally generated. This can be found in the actual class declaration at the beginning of the code:

    self.contour = create_noisy_circle_curve(1, .2, 30)

    Every Trunk, Branch, and NonBranch object have a their own unique curve associated with them. However this curve is only applied in the grow_one(o) section:

    if 'Trunk' in o.name or 'Branch' in o.name:
        nproduce SetContour(o.contour)
    else:
        reset_contour()

    This code targets every Trunk, Branch, and NonBranch object and sets their own curve when growing them. Whenever any other object is passed to the grow_object(o) function the call to reset_contour() sets the contour back to a perfect circle to ensure that no curves overlap.

    The create_noisy_circle() function is included in the helper.py file. It works by creating a circle out of a given number of points, and then moving those points in the x and y direction according to a given noise factor. The function has two required parameters and two optional parameters. The two required parameters are radius and noise_factor. The radius determines the radius of the generated circle. The noise_factor is used to set a range in which random points will be generated. The points making up the circle will then be moved a random amount in that range. This can be seen in the main for loop in the function:

    for angle in t:
        # Base circle points
        x = radius * cos(angle)
        y = radius * sin(angle)
    
        # Add noise
        noise_x = uniform(-noise_factor, noise_factor)
        noise_y = uniform(-noise_factor, noise_factor)
    
        noisy_x = x + noise_x
        noisy_y = y + noise_y
    
        points.append((noisy_x, noisy_y))

    The two optional parameters for the function are num_points and seed. num_points is used to determine how many points make up the circle. If no value is given, it defaults to 100. seed is used to set the randomness of the circles. If a value is given, the random.seed is set to that value. For this model, seeds were not used.

  5. Changing Tertiary Branch Curves:

    Every tertiary branch on the model follows a unique curve. This curve is generated with the create_bezier_curve() function which can be found in the helper.py file. The function works by generating four points to be used as guide points for the Bézier curve. There is some control code to make sure that the x points are random but are still generated in a linear fashion.

    The function takes four optional parameters: num_control_points, x_range, y_range, and seed_val. num_control_points defaults to four, the standard for Bézier curves. x_range and y_range default to a tuples designating the ranges: (0,10) and (-2,2) respectively. seed_val defaults to None, however in the actual model time.time() is used as the seed.

    The actual designation of the curves for the tertiary branches can be seen in the bud(t) section:

    if 'NonTrunk' in new_object.name:
        import time
        curve = create_bezier_curve(seed_val=time.time())
        nproduce [SetGuide(curve, new_object.max_length)

    A curve is generated with the create_bezier_curve() and it is used as the guide for the SetGuide function provided by L-Py.

  6. Changing color ID system:

    Every object on the tree has its own unique color code to act as an ID. The implementation of this can be seen in the grow_object(o) section:

    r, g, b = o.color
    nproduce SetColor(r, g, b)
    
    smallest_color = [r, g, b].index(min([r, g, b]))
    o.color[smallest_color] += 1

    This works by finding the smallest value of the objects RGB code and incrementing it by one. This is to avoid any kind of error where a color value is greater than 255. To get rid of the color IDs, simply comment out the second two lines of the code above.

  7. Changing Apple and Leaf ratio:

    The generation of apples and leaves is random. If something is to grow off of the bud, there is a 90% chance it will be a leaf, and a 10% chance it will be an apple. These percentages are set in the Spur class declaration:

    # From Spur class
    def create_branch(self):
        if self.num_leaves < self.max_leaves:
            self.num_leaves += 1
            if rd.random()<0.9:
                new_ob = Leaf(copy_from = self.prototype_dict['leaf'])
            else:
                new_ob = Apple(copy_from = self.prototype_dict['apple'])
        else:
            new_ob = None
    
        return new_ob

Features

  • Pruning: Remove unwanted branches to simulate pruning.
  • Branch Tying: Simulate branches being tied down to mimic different orchard architectures.
  • Model Class Types: The model generated is built with classes of different material type.

Gallery

media/envy_model.png

Example of a labelled, pruned and tied envy tree system using TreeSim_Lpy

media/ufo.png

Example of a labelled, pruned and tied UFO tree system using TreeSim_Lpy

Contact

For any questions or issues, please contact us through GitHub Issues.

Help and Support

Please open an Issue if you need support or you run into any error (Installation, Runtime, etc.). We'll try to resolve it as soon as possible.

Citations

    1. Boudon, T. Cokelaer, C. Pradal, P. Prusinkiewicz and C. Godin, L-Py: an L-system simulation framework for modeling plant architecture development based on a dynamic language, Frontiers in Plant Science, 30 May 2012.

About

Grows trees including the features of tying and pruning based on pre-defined rules. Generates 3D models of these trees

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages