cjcormier / roboscaffold_sim Goto Github PK

NOTE: The simulation has been ported to Kotlin and this repository is deprecated.

Originally there was going to be support for running the kotlin simulation via this UI, but the UI was ported to JavaFX/TornadoFX. Thus some of the functionality hinted at by the UI is non-functional (namely drawing the structure and changing simulation language, these options are left for posterity).

#Overview This is the repository for the simulation of the 2017-2018 Robotic Scaffolding Major Qualifying Project (MQP) at WPI.

The goal Robotic Scaffolding MQP is attempt to create a system that allows for autonomous creations of two dimensional structures out of building material. This is done by two sub-systems, the scaffolding network and the builder robot, working together.

The scaffolding network is composed of blocks that all connected via other scaffolding blocks. The scaffolding blocks can communicate to all other connected blocks, in the physical demo this is achieved via a CAN bus. The scaffolding blocks can also talk to the a builder robot that is directly above it. This is achieved via changing an LED on the top face of the block. The scaffolding blocks have no actuation in of themselves.

The builder robots drive around on top of the the scaffolding network, receiving instructions from the scaffolding blocks below it. These instructions tell the robot where to drive and which blocks (scaffolding or building) to pick up and where to place them. The builder robot is assumed to have no concept of its location and can only hold a single block of any type. (While there is some error checking using the held block, this is to help debug different algorithms and should not be used in the algorithms themselves).

These two systems work together to place building blocks in the shape of the target.

#Requirements The simulation requires python version 3.6 or higher and tk/tcl version 8.5 or higher installed. To see if the correct version of tk is installed run tk_check.py. If the wrong version is installed, follow the instructions here.

#Applications and Tests There are two applications in this project, the basic_simulation_visualizer.py and strategy_profiling.py applications. Both of these applications are in the root folder of this project.

NOTE: All applications and tests must be started from the project root folder, otherwise the libraries will not be found correctly. (For example to run test create_point_test.py", run "python3.6 roboscaffold_som/tests/create_point.py).

##basic_simulation_visualier.py This is the main application for this project. The application consists of two views: the creation views and the simulation view.

The creation view allows for specifying the target structure for the simulation. There are currently three ways to specify a structure: choosing a pre-made structure, choosing a dimension-value pair, and loading a file. The ability to create a structure block by block was planned but was not implemented before porting the project.

The creation view allows for saving the defined structure and running the simulation in the same window or a new one. When running the simulation, one of multiple strategies can be selected. The strategies are all variations of the same basic strategy, but with various vertical offsets.

For any specified size there are multiple ways to number the possible structures. The simplest way is to number them in binary with each digit representing whether a coordinate is in structure or not. The method used here is similar but requires a block to exist in the first row and column this is to reduce repetition. Currently it is not possible to see the max value for the given dimension in the creation view as this feature was not added before the port.

The simulation view allows the user the run the simulation step by step, or play it at the given frames per second (FPS). Only the first 1000 frames are loaded initially, but more can be load after. Statistics about the currently loaded frames are also availible, as well as the ability to save the run.

This application tests several strategies and determines which are better according to different metrics. The strategies tested are the Basic Spine, Offset Spine, Centroid, and Longest Spine Strategies. These strategies are run on all valid structures of that fit in a square (the dimension of which can be set in the file).

There are two metrics that the strategies are tested on the number of robot moves and number of scaffolding updates. The number of robot moves is more impactful as it takes much more time for a robot to more than the scaffolding to update.

NOTE: Be aware that the runtime of strategy profiling is O(2^(n^2)) where n is the dimension of the square because of the number of possible structures. Due to this, it is not recommended to run it for more than a dimension size of 5 (which may take several hours).