All these ideas build on work discussed in class, and are least partially novel. The Litus and Lein papers were based on work done in this class.
- Joint recharging decisions: show that robots can achieve more discounted work if they jointly decide their battery threshold to go recharge, compared to any fixed threshold. See [Berenz:2012] for a single robot model.
- Distributed simultaneous task allocation & navigation: Can you modify the LOST framework to do optimal load balancing between multiple food sources?
- Add additional information to the crumb cost label.
- Use a FASR-type method to change a robot's choice of which trail to use.
- Adaptive spatial interference reduction: can you get a robot population to adaptively switch between individual foraging and bucket-brigading depending on the current spatial interference? See [Lein:2009] for a partial solution.
- FLOST: Investigate adding flocking behaviour to LOST robots, in addition to obstacle avoidance. Can you reduce spatial interference and terrain coverage to find better trails?
- Daydream LOST: Can a robot build a world model online, then simulate LOST in the model to find efficient paths without spending all that energy in the real world?
- Reinforcement learning for local obstacle avoidance with cooperating (and possibly uncooperating?) dynamic obstacles. Can you learn a control policy from sensor data that beats RVO?
- Ordering strategies in the Frugal Feeding Problem: devise and evaluate distributed heuristics for choosing the order of individual rendezvous in the FFP. See [Litus:2007]
- Pheromone Robotics: In [Payton:2001] Pheromone robots deploy themselves as stationary graph nodes. If we allow nodes to continue moving, could we usefully combine Payton's exploration functionality with formation control?
- Robot sheepdog teams: devise novel control algorithms for multiple robot sheepdogs to move a flock around an obstacle course. Ideally, the methods would scale to any number of sheepdogs >= 3 and any number of flock creatures.
- Multi-robot area coverage with fractal trajectories: extend [Sadat:2015] to multiple robots.
- Decentralize Amazon's Kiva robots. Devise a distributed algorithm that addresses the Kiva planning problem, and competes with a centralized planning approach. Video. This application is very important with lots of players in the space.
- Cooperative manipulation: Can you extend Lily's recent work on light-field controlled robots to transport large objects? See Push code at GitHub.
- [Berenz:2012] Autonomous battery management for mobile robots based on risk and gain assessment. Berenz, V., Tanaka, F. & Suzuki, K. Artif Intell Rev (2012) 37: 217. doi:10.1007/s10462-011-9227-9. ONLINE
- [Lein:2009] Adapting to non-uniform resource distributions in robotic swarm foraging through work-site relocation. Adam Lein, Richard Vaughan.
Proc. IEEE Int. Conf. on Intelligent Robots and Systems (IROS'09), St. Loius, MO October 2009. PDF
- [Litus:2007] The Frugal Feeding Problem: Energy-efficient, multi-robot, multi-place rendezvous.Yaroslav Litus, Richard Vaughan, Pawel Zebrowski.
Proc. IEEE Int. Conf. on Robotics and Automation, April 2007. PDF
- [Payton:2001] David Payton , Regina Estkowski , Mike Howard Compound Behaviors in Pheromone Robotics Robotics and Autonomous Systems, 2001, 44, pp.229--240. PDF
- [Sadat:2015] Fractal Trajectories for Online Non-Uniform Aerial Coverage
Seyed Abbas Sadat, Jens Wawerla, Richard Vaughan.
Proc. IEEE Int. Conf. on Robotics and Automation (ICRA'15), Seattle, USA May 2015 PDF