Update GSoC24 project descriptions (#838)

events/2024-google-summer-of-code/index.markdown

@@ -17,35 +17,82 @@ More information about the MoveIt GSoC project applications [tips at the bottom
 ## Project Ideas
 
 ### Better Simulation Support
-Experiment with alternative simulation backends like MuJoCo, O3DE, Gazebo while improving interoperability with MoveIt.
-Investigate new use cases and best practices, and document your findings in a well-structured review page including updated tutorials and demo implementations.
+Experiment with alternative simulation backends like [MuJoCo](https://mujoco.org/), [O3DE](https://o3de.org/), [Gazebo](https://gazebosim.org/home) while improving interoperability with MoveIt.
+Investigate new use cases and best practices, and document your findings in a well-structured review page including updated tutorials and demo implementations. An example application for showcasing improved simulation support could be running an admittance controller for solving a peg-in-hole problem.
+
+**Mentors**: [David](https://github.com/dyackzan), [Henning](https://github.com/henningkayser)\
+**Project Size**: 175 hours\
+**Difficulty**: Medium\
+**Skills Required**: ROS 2 and MoveIt basics, C++ or Python, experience with at least one of the named simulation backends\
+**Outcome**: A repository providing simulated and documented examples for example use cases. Improved tutorials on setting up supported simulation backends, and a guide that compares the features for various applications.
+
 
-Mentor: [David](https://github.com/dyackzan), [Henning](https://github.com/henningkayser)
 
 ### Improved Collision Avoidance
-Integrate alternative collision checking backends like HPP-FCL, or research SDF-based approaches for improving MoveIt's collision checking capabilities.
+Integrate alternative collision checking backends like [HPP-FCL](https://github.com/humanoid-path-planner/hpp-fcl), or research SDF-based approaches for improving MoveIt's collision checking capabilities.
 Experiment with new use cases like continuous collision detection or online collision avoidance.
 
-Mentor: [Paul](https://github.com/pac48), [Henning](https://github.com/henningkayser)
+Some example ideas include:
+  - Support HPP-FCL as collision checking backend in MoveIt (see [#1435](https://github.com/ros-planning/moveit2/issues/1435), [#3167](https://github.com/ros-planning/moveit/pull/3167), [#1521](https://github.com/ros-planning/moveit2/pull/1521))
+  - Update or reimplement MoveIt’s discrete SDF approach, apply with Servo or STOMP
+  - Implement a sphere-based or convex mesh decomposition of the robot geometry for accelerated collision checking
+
+**Mentors**: [Paul](https://github.com/pac48), [Henning](https://github.com/henningkayser)\
+**Project Size**: 350 hours\
+**Difficulty**: Medium/Hard\
+**Skills Required**: C++, familiarity with ROS 2, knowledge of 3D representations, experience with FCL/Bullet or similar libraries\
+**Skills Preferred**: knowledge of convex decomposition and SDF, hardware acceleration\
+**Outcome**: Updates to MoveIt’s collision checking interface and plugins. Implement and improve tooling for preprocessing and optimization of the robot’s geometry. Benchmark collision checking use cases, aim to achieve a significant improvement in collision checking performance.
+
 
 ### Drake Integration Experiments
-Review and understand the different motion planning and control capabilities of Drake and MoveIt.
-Document options for making both frameworks interoperate in solving particular use cases.
-Select at least one use case in MoveIt that could be improved or enabled by using features provided in Drake, and implement a demo application.
+MoveIt is a motion planning framework that is well-integrated with ROS, whereas [Drake](https://drake.mit.edu/) is a model-based planning and control framework with state-of-the-art optimization capabilities. This project seeks to combine the strengths of both these approaches by exploring the use of Drake within MoveIt’s planning pipeline or hybrid planning framework for tasks such as motion planning, trajectory optimization, and/or online control. The ideal candidate will have familiarity with at least one of these two software tools, and in their proposal would describe their specific ideas towards the desired outcomes.
+
+Some example ideas include:
+  - Creating a MoveIt motion planning plugin that uses Drake’s trajectory optimization based planners, such as direct transcription and direct collocation.
+  - Creating a MoveIt planning response adapter that uses TOPPRA for time-optimal parameterization.
+  - Using Drake’s dynamical modeling capabilities for online control, such as Model Predictive Control (MPC), as a local planner in MoveIt.
+
+
+**Mentors**: [Sebastian Jahr](https://github.com/sjahr), [Sebastian Castro](https://github.com/sea-bass)\
+**Project Size**: 350 hours\
+**Difficulty**: Hard\
+**Skills Required**: Experience with C++, ROS 2, optimal planning and control, and at least one of MoveIt or Drake\
+**Outcome**: Releasing one or more experimental MoveIt capabilities that utilize Drake for motion planning, trajectory optimization, and/or online control. This should include supporting documentation and examples, as well as sharing a general methodology for using Drake with MoveIt.
 
-Mentor: [Sebastian Jahr](https://github.com/sjahr), [Sebastian Castro](https://github.com/sea-bass)
 
 ### Support Closed-chain Kinematics
-Investigate use cases for closed-chain kinematics planning problems and MoveIt's current limitations in solving them.
-Select at least one demo application to implement and demo while improve MoveIt's planning capabilities.
+MoveIt’s kinematic representation is modeled as a tree which implies an open kinematic chain. However, robots are supposed to operate in an environment and perform manipulation tasks that may involve closed-chain kinematics or dynamics: e.g. opening and closing doors and bi-manual object manipulation. While these can be approximated using open chain FK and IK, these techniques may introduce undesired inaccuracies which translate to poor trajectory and controller execution. This project is about researching and implementing methods for improving the representation of the robot’s kinematic or of kinematic constraints in the scene, and to implement optimized kinematic solvers for computing them.
+
+**Mentor**: [Tyler](https://github.com/tylerjw), [Henning](https://github.com/henningkayser)\
+**Project Size**: 350 hours\
+**Skills Required**: Research background in robotics, expert knowledge of kinematic representations of robots\
+**Difficulty**: Hard\
+**Outcome:**
+  - Example configurations and use cases for applications with closed-chain kinematic problems
+  - Documentation and tutorials on solving motion planning problems involving closed-chain kinematics
 
-Mentor: [Tyler](https://github.com/tylerjw), [Henning](https://github.com/henningkayser)
 
 ### Zenoh Support & Benchmarking
-Enable and test the new ROS 2 middelware backend Zenoh running MoveIt and auxiliary runtime dependencies.
-Set up a simple benchmark for testing use cases like visual servoing for stressing the middleware, and compare the results with the existing DDS implementations.
+[rmw_zenoh](https://github.com/ros2/rmw_zenoh) is the exciting new middleware in ROS 2 in Jazzy. MoveIt is a complex project that depends on many ROS 2 features. We want to understand how using rmw_zenoh impacts systems using MoveIt. We also wish to document the configuration changes needed to make rmw_zenoh work well with MoveIt.
+To do well in this project, you’ll need to be experienced in configuring and using MoveIt and ROS. A good primer for this project would be to work through the MoveIt, ROS, and DDS tutorials to configure the various DDS middlewares and understand their impacts on the robot system. To make a fair comparison and write helpful docs, you must understand how to configure ROS with the various middlewares.
+Doing this project will allow you to contribute to the newest development in ROS and enable others who build MoveIt-based robot projects to make the correct middleware choices for their projects.
+
+
+**Mentor**: [Henning](https://github.com/henningkayser), [Tyler](https://github.com/tylerjw)\
+**Project Size**: 90 hours\
+**Skills Required**: ROS 2 experience\
+**Difficulty**: Medium\
+**Outcome**:
+  - Documentation on Zenoh configuration and use with MoveIt contributed to the tutorial site.
+  - Find and report bugs in rmw_zenoh when used with MoveIt
+  - Benchmark impacts on message latency
+  - Benchmark impacts on large messages such as point clouds
+  - Benchmark how the number of topics and participants impacts performance
+  - Contribute changes to MoveIt and ROS where applicable to support Zenoh users
+  - Contribute CI jobs to test MoveIt with each ROS middleware
+  - Benchmark the use of the zenoh-dds bridge in comparison to rmw_zenoh
 
-Mentor: [Henning](https://github.com/henningkayser), [Tyler](https://github.com/tylerjw)
 
 
 ## Tips for writing a successful Google Summer of Code application for MoveIt