Project Ideas

1) Have an idea? Add it!

2) Want to joint a group? Contact them!

3) Group leader updates the names on the project

*Based on the size of the class there should be approximately 5 groups for final projects. There is now a column where Mike lets you know that the basic concept of the project is approved. You will still need to get Mike and undergraduate reviewers to approve the specifics of your proposed algorithms / implementations.

Member Name(s)	Project Idea	Description	Mike Approves
Can Erdogan, Azfar Aziz, Matthew Dutton	"Make a robot that can design (and build) a rube goldberg machine"	If you have ever designed and/or built a Rube Golderberg machine, you would know how challenging it can be for many different reasons. The primary focus of this project is to appreciate and resolve the underlying planning problem. An example can be moving a ball from one end of the "scene" to another, using a set of tools, such as blocks, springs, strings, balloons, dominoes, etc. We will work in the simulation environment, probably for the length of the semester. However, we definitely want to extend this work to the real world (it would be so cool!) after the class. So, we hope this will become a long-term project. Also, if possible, we would like to perform learning, as much as possible. As we finalize our group, we can start discussing any ideas you come up with and formulate the problem together.	X
Arjun Menon, Stephen Motter, Yunfei Bai, Patrick Dillon	"Multi-task manipulation"	The goal of this project is to simulate the motion to accomplish multiple manipulation tasks simultaneously. The input of the method is the description of tasks along the time line, such as moving an object from one place to another during a time interval. And the output is a sequence of animation accomplishing all the tasks. The manipulation is a redundant system that means the character can have multiple configurations even though carrying out the same job. And this makes it possible for the character to achieve different tasks simultaneously. Instead of giving different tasks diverse priorities as in literature, we treat all the tasks equally and try to find a motion that can satisfy all of the tasks by solving an optimization problem. The motion of upper body and lower body are handled separately. For lower body, it will determine the character's position, orientation and height. The prerecorded data such as walking, turning, bending the waist, squatting, etc, can be straightly used. As for the upper body, the motion is simulated to accomplish a specific task. A motion graph is defined for the upper body which contains the end effectors that can be used and the transitions between each end effectors. In our manipulation graph, we have seven end effectors which are left hand, right hand, left arm, right arm, left shoulder, right shoulder and both hands. Constraints are also defined for manipulation graph which will determine which end effector can be used for a task. Here is an example showing how this approach works. A task is defined as holding a coffee cup while picking up a backpack from the ground, then walking through the room to open the door with the coffee cup and backpack in hands. The algorithm will first determine which end effector to use, and then apply recorded data for lower body motion, while searching for proper joints' torques to control the upper body by solving the optimization problem to complete all tasks. The simulated result may be the character holding the cup by left hand while picking up the backpack on right shoulder and then using right hand to open the door.	X
John Turgeson, Taylor Weiss, Joseph Rogers	Path planning for object acquisition, analysis and distribution	Our mission is to select an item from a collection of various object, determine the nature of the item and place each into a more desirable location. In an attempt to acquire a full visualization of the object it will become necessary to set the object on a surface and grasp at an angle allowing features that were previously hidden to be registered. The object can then be placed in a desired location based on it's structure. Applications include sorting of parts in an industrial setting, dishes, clothing and toys in a domestic setting, or debris analysis during post disaster events. Planning will become necessary to move the object in the workspace for visualization, maintaining an unobstructed line of sight to the object, and proper gripping location for analysis of the second half of the object.	X
Adityan Srinivasan, Munzir Zafar, Joan Davis	Sword-fighting for Krang	We aim to teach a humanoid robot sword-fighting. The robot we intend to use is Krang, in Mike's Golem lab. Our goal is to develop a simulation for this using the software codebase for dynamic visualization currently being developed by GVU at Georgia Tech. We would like to see this extended to Krang after developing the simulation this semester.	X
Sameer Ansari, Billy Gallagher, Kyel Ok, William Sica	Simultaneous Planning Localization And Mapping For UAVs	The goal of the project is to implement a planning technique into SLAM for UAV's, using a local planner to direct the path of the UAV towards the global exploration goal. This work is motivated to improve SLAM performance in uncertain or rapidly changing environments, such as search and rescue missions in a forest.	X