Logistics

This assignment is due on Nov 6th, 2026 at 10pm. As with all assignments, you will have the opportunity to revise this assignment based on the feedback that you get.

Collaboration policy

For this assignment, you may discuss the assignment at a high level with other students, however you must type up all of the code yourself. (It is therefore unexpected for two code submissions to be completely identical.) You can discuss the assignment with Anya, our course staff, or any lab assistants.

You should list any student or course staff with whom you discussed the assignment and the manner of discussion in your README.md file.

AI policy

As discussed in the syllabus, there are both acceptable and unacceptable uses of genAI in this class. Regardless of whether you use any genAI, you need to complete the AI Log in the README.md to satisfy the advanced requirement.

If you choose not to use genAI at all, you may state that for your AI Log (and I fully support this choice). I’d be interested in your reasons, though it’s not required to write them in the log for every assignment.

If you do choose to use genAI, you should log it in the AI Log section of the README in the following format:

  • Date: Summary of use and assessment of usefulness

For example, a hypothetical log could be:

  • 8/25/26: I asked Gemini how to do default values for a constructor and it helped me figure it out
  • 8/26/26: I couldn’t figure out the problem with the ApplyLinearSpringForces method and Gemini was completely useless
  • 8/26/26: I googled the problem with the ApplyLinearSpringForces and the AI summary was helpful

This log requirement is both to give me an idea of how useful genAI is in this context and to make sure that your learning isn’t being hampered. As mentioned in the hypothetical log, you should include any genAI resources that you use, even the AI summary shown in search results or GitHub Copilot (which shows up as autocomplete in VSCode now).

Goal

The goal of this assignment is to demonstrate your understanding of digital evolution and work within a full artificial life research software project. You’ll be expanding a digital evolution project that runs in the web browser and command-line using the Symbulation platform, which is built with the Empirical and SGP-Lite frameworks.

Setup

You’ll be working in your own copy of the Symbulation software for this project. I’ve provided that copy for each of you along with a couple of scripts to help make things easier. You are welcome to tweak other aspects of the software, but there is a lot so definitely don’t try to look through it all!

Symbulation has extensive documentation along with Empirical’s documentation, and I recommend you take advantage of the search functionality of both.

  1. Go to our 361-F26 Organization and find the Symbulation repository for your account. Use git clone --recurse-submodules [URL] as usual.

  2. Then, go to the emsdk folder and install and activate emsdk: 
    cd emsdk
./emsdk install 3.1.49
./emsdk activate 3.1.49
  3. Then source emsdk. You’ll need to do this command anytime you open a new Terminal, so if you are getting an error about emcc not existing, you probably forgot to do this: 
    source ./emsdk_env.sh
cd ..
  4. Verify that everything is set up correctly by going back to the top level of your repository and running the compile-run-web.sh file (there will be a lot of warnings, don’t worry about them): 
    ./compile-run-web.sh

Supplemental info

As you are working on this project, you may find the supplemental material section F of this paper a helpful reference for what operations are available and what they do, as well as the SGP-Lite lab.

You may also find this Symbulation Architecture Guide helpful.

Assignment Specification

For this assignment, you will expand on the Symbulation starter code that is provided by implementing additional tasks and conducting experiments.

Tasks

First, you should add at least two tasks to the predefined_tasks set in source/sgp_mode/tasks/LogicTaskSet.h. You will also need to edit default-config-jsons/diff-reward-env.json to include your new tasks. You may choose to remove any of the existing nine if you wish as well, or you can just add on.

Then, you need to edit web/SGPSymAnimate.h task_map and matchColor (or other places if you prefer) to visually show your tasks and report in your README.md which tasks you implemented and how you are visually representing them. You should be sure that they show up within a reasonable amount of time in the viewer.

Basic Experiment

Now that you have your new tasks implemented, you should explore one of the existing settings/options and find two values of the setting that lead to your new tasks being more or less likely to evolve (you don’t need to conduct rigourous statistics, but there should be a noticable difference between experiments). You can conduct your initial exploration with the web GUI by removing the relevant ExcludeGroup or by running in the command line and looking at the Data/Tasks_data.csv file.

Once you’ve found a setting/option that changes the evolution of your task based on its value, you should run experiments with at least 5 different random seeds for each and report your results in the core section of your README. You may either just include the raw numbers or you may plot your data using whatever plotting software you are comfortable with. The folder in the starter code stats_scripts contains some potentially useful scripts, including simple_repeat.py and munge_data.py as well as some (very bad) R scripts that I used in the past (please don’t judge the terrible R-style). The Collecting and Analyzing Quickstart Guides should also be helpful.

Here are some ideas for settings/options that are likely to influence the evolution of your tasks at different values (remember that you can see all possible settings by looking in the files source/sgp_mode/SGPConfigSetup.h and source/ConfigSetup.h):

  • HOST_REPRO_RES
  • WORLD_WIDTH
  • WORLD_HEIGHT
  • HOST_ONLY_FIRST_TASK_CREDIT
  • SGP_MUT_PER_BIT_RATE
  • Removing some of the other tasks from default-config-jsons/diff-reward-env.json

Advanced Expansion

For the advanced portion of the project, you will further expand on Symbulation in a direction of your choosing and then conduct another experiment to determine the effect of that expansion (it’s fine if there isn’t one this time though).

Here are your options for expansions:

  • Create a custom event type (separate from a TaskValueEvent)
  • Create a custom spatial structure (which will be a represented as an adjacency list or matrix, examples are in source/test/data/chain-edges.csv etc)
  • Create three additional tasks such that you have at least five new tasks that build up in an interesting/meaningful way (and have them visually represented in your GUI)
  • Turn on symbionts via START_MOI set to 1 and explore something beyond what the papers did (if you choose this option, you should check with me that your focus is complex enough)

You should again run the software with at least 5 different random seeds each for at least two treatments (i.e. with and without your expansion for example), save the relevant data from Data and plot relevant values with your choice of plotting software. You should additionally include at least 1 paragraph discussing the meaning of your results.

Assessment

The core requirements are:

  • Run and display an updated digital evolution system when the compile-run-web.sh script is run, using the web-browser, SignalGP-Lite, Symbulation, and Empirical libraries
  • Have at least two new tasks that organisms evolve to perform within a reasonable amount of time (running your code for about 10 minutes) which are visually represented in your GUI (you can choose whether to keep the default Logic 9 tasks, but you must implement two different tasks)
  • Specify in the Core section of your README.md what your new tasks are and how they are visually represented in your GUI
  • Conduct a basic experiment with your new tasks and report in your Core section of your README on your results (at least 1 paragraph, results must show a noticable difference between settings)

The advanced requirements are:

  • Complete the core requirements (including the basic experiment)
  • Implement one of the advanced expansions and experiments
  • Specify in the Advanced section of your README.md what expansion you chose and a report of your results (at least 1 paragraph with a plot of your data)

Submitting

Submit your code to GitHub with the usual add/commit/push steps. Then tag your most recent commit with the following:

git tag -a initial_submission -m "initial submission"
git push origin initial_submission

Getting Help

Remember that you can and should ask for help on everything from annoying C++ bugs to Git to “is this a reasonable expansion to implement?” The CampusWire is probably the best place to ask, as long as you don’t have to share too much of your code to do so, because you’ll get faster responses on average. You are also welcome to come to drop-in hours and schedule a one-on-one meeting if that works better for you.

Acknowledgements

Symbulation has many additional contributors, here is the full citation: Vostinar, A., Johnson-Horrigan, K., Lalejini, A., Dolson, E., Hillman-Emelianoff, E., Shea, J., Kelley, D., Khalili, A. T., Welch, P., Cameron, A., Orion Hafner, L., Dorchen, S., Gormely, C., renztim, Ofria, C., Qin, Y., & mkmmnn. (2026). anyaevostinar/SymbulationEmp: ComplexFeatures (Version v1.0) [Computer software]. Zenodo. https://doi.org/10.5281/zenodo.20732915