# CSCE 689-604: Special Topics in Multi-Robot Systems

## Project Ideas

These are examples of generic ideas, which give a feel for the expected scope of the project:

- Implement a variation of the Kube-style minimalist box-pushing, but in which robots can form chains behind one another in order to increase pushing strength. (Suggested by Shruva Nath.)
- Consider a variation of the Holland/Melhuish clustering along with Kube-style minimalist box-pushing wherein boxes can be pushed either individually or collectively, and robots can employ different mixtures of strategies for collecting together and pushing.
- Design and implement a variation of the Holland/Melhuish clustering experiments in order to bias clustering toward some pre-defined shape.
- Design and implement a variation of the Holland/Melhuish clustering experiments in order to model ant corpse-clustering more precisely.
- Consider a multi-robot charging problem, in which you have more robots than
chargers. Have the robots do some other task (
*e.g.,*a surveillance task) and sharing charging stations as necessary for long-term autonomy. (See the paper called "Basic Cycles, Utility and Opportunism in Self-Sufficient Robots" by David McFarland and Emmet Spier.) - Implement and evaluate a method for incremental deployment using network communication. (See this paper entitled "Incremental Multi-Robot Deployment for Line-of-Sight Chains Using Only Radio Signal Strength" -- code can be provided as a starting point.)
- Implement several path planning approaches to give different quality
estimates of the expected path length (
*e.g.*, straight-line, vs. full path on a high-resolution configuration space representation with navigation constraints) and assess the degree to which they correlate with time spent navigating. Then assess the implications for algorithms that assign navigation tasks to groups of robots. - Read this
paper on sub-linear time estimation of swarm properties using local messaging. These are
three possible extensions, each suitable for a class project:

(a.) propose an extension to estimate density? (What about variation in density?). Implement and evaluate your idea. (Some ideas and code to start with can be provided.)

(b.) can the performance be improved for swarm sizes? (This is described in the future-work, via smoothing the boundary properties, constructing a convex hull,*etc.*)

(c.) how can the idea be extended to 3D?

- Implement and test the auction-based task-allocation method for box-pushing algorithm in the paper by Gerkey and Mataric': Sold! Auction-based methods for multi-robot coordination.
- Propose, implement and evalutate a distributed algorithm for scaling 802.11 or 802.15.4 transmit power to as to keep local communication to a fixed
*k*neighbors. (Look here for starting point on how setting can be performed.) - Propose and evaluate an extension of the idea in Hummel, Taylor and Hover's "Numerical Optimization of Generative Network Parameters" ASME International Mechanical Engineering Congress and Exposition, 2010, but for the multi-robot domain. In other words, are there principles we can use to produce networks (or control policies which result in networks) with known, useful robustness properties.
- Implement and test the algorithm you proposed in your Task-Allocation assignment.