This year's All-Energy conference was cancelled, but that didn't stop the Victorian branch of the Australian Institute of Energy from holding their own postgraduate energy student poster competition, held virtually. And this year, the winner was our very own Saman Ahmadi, with his presentation on Planning Energy Efficient Routes for Electric Vehicles. Congratulations, Saman!
Notifications are out for AAAI 20201 submissions, the last (but not least) venue in our 2020 conference season. Acceptance was extremely competitive this year, with an acceptance rate of 21% (over an incredible pool of over 9000 submissions). With this in mind, we had an impressive showing this year, with 7 papers accepted across the group:
- Eli Boyarski, Ariel Felner, Pierre Le Bodic, Daniel Harabor, Peter J. Stuckey, and Sven Koenig. f-aware conflict prioritization & improved heuristics for conflict-based search.
- Emir Demirovic and Peter J. Stuckey. Optimal decision trees for nonlinear metrics.
- Jo Devriendt, Stephan Gocht, Emir Demirovic, Peter J. Stuckey, and Jakob Nordstr̈om. Cutting to the core of pseudo-boolean optimization: Combining core-guided search with cutting planes reasoning.
- Zhe Chen, Daniel Harabor, Jiaoyang Li, and Peter J. Stuckey. Symmetry breaking for k-robust multi-agent path finding.
- Alexey Ignatiev, Edward Lam, Peter J. Stuckey, and Joao Marques-Silva. A scalable two stage approach to computing optimal decision sets.
- Graeme Gange and Jarrod Knibbe. Optimising automatic calibration of electric muscle stimulation.
- Saman Ahmadi, Guido Tack, Daniel Harabor and Phil Kilby. A Fast Exact Algorithm for the Resource Constrained Shortest Path Problem.
A joint team of researchers from Monash and The University of Southern California have taken top of the final leaderboard in the 2020 Flatland Challenge!
Organised by Swiss Rail and a consortium of other industry partners, this event asks for effective new solutions to challenging rail planning problems. The competition this year was extremely fierce: the top two teams traded places several times in the final days and the winning entry was submitted just 15 minutes before the deadline!
Participating team members are (in alphabetical order): Shao-Hung Chan (USC), Zhe Chen (Monash), Jiaoyang Li (USC), Hang Ma (Simon Fraser University),Yi Zheng (USC). The team was advised/coached by Daniel Harabor (Monash), Sven Koenig (USC) and Peter Stuckey (Monash).
The newest version of MiniZinc is released, and available at MiniZinc.org. As usual, it comes with a bevy of minor enhancements and fixes. Give it a try!
The most exciting feature of the 2.5 family is the ability to extend enumerated types with Enum constructors, so you can add extra elements to Enums without coercing back to integers (and losing index-set checking).
CP this year was closely followed by CPAIOR. Originally to be held in Vienna colocated with SOCS, CPAIOR was likewise delayed, and run fully online. Being run on a Vienna schedule, this meant a return to a nocturnal lifestyle for our intrepid real-time attendees. Fortunately, all sessions were recorded for the diurnal among us.
The group presented three papers, and Peter Stuckey gave an invited talk on some of our recent work applying optimization techniques to multi-agent pathfinding. As usual, the conference covered an interesting mix of work on underlying optimization techniques, and on specific applications. Congratulations to the organizers for putting on a successful virtual edition.
The 26th International Conference on Principles and Practice of Constraint Programming (CP 2020) was over Sep. 7-11th 2020. Nominally organized in Louvain-la-Neuve Belgium, it was run as a fully-online virtual conference. The time difference made for an interesting week of fatigue for those of us attempting to attend the conference in real-time.
As in previous years, our group demonstrated an impressive presence at the conference, presenting 10 papers; Jinqiang Yu, Alexey Ignatiev, Peter Stuckey and Pierre Le Bodic were awarded best paper in the CP/ML track for "Computing Optimal Decision Sets with SAT". Congratulations, all!
The conference concluded on sombre note, with a memorial session for Christian Schulte, who passed away from a sudden illness early this year. Christian was a massively influential figure in the field, both for his work on the Oz and Gecode constraint programming systems, and in making the community a welcoming place for newcomers.
The International Symposium on Combinatorial Search (SoCS) is an annual meeting of researchers and practitioners working in the area of Heuristic Search. This year marks the 13th edition of the conference with Daniel Harabor (Monash University) and Mauro Vallati (University of Huddersfield, UK) being co-chairs. Originally scheduled to take place in Vienna, Austria, SoCS 2020 was this year entirely online. The program included: 27 technical talks in five streaming sessions. Two wonderful keynotes, from Rina Dechter and Peter Stuckey. For the first time ever at SoCS, an entire day of Master Class sessions, where recognised experts tell us about recent developments in Search. Community participation took place via a dedicated forum and Discord server. The whole thing was regarded by the community as a big success! Check out some of the lessons learned in our post-conference article, soon appearing in AI Magazine!
If you ever wanted to know more about using MiniZinc and constraint solving as a tool for decision making in practice, we’ve now got you covered! Our very own Mark Wallace has released a new book, ‘Building Decision Support Systems (using MiniZinc)’.
In addition to demonstrating how to model a range of optimization problems (with copious examples in MiniZinc), the book also goes deeper into some of the risks and challenges involved in using optimization as a tool in practice. Writing any book is a mammoth undertaking, so congratulations Mark!
Hot on the heels of the Amazon Research Award, Graeme has also received a Google Faculty Award, for improving parallelism in nogood learning constraint solvers. Nogoood learning (or lazy clause generation) solvers speed up search by learning new constraints from each failure, to prune similar parts of the search space. But combining this with parallelism is tricky: sharing learnt clauses requires (expensive) communication between workers, but without sharing workers waste effort duplicating search.
Graeme Gange has received an Amazon Research Award for ‘Robust Prioritized Multi-agent Pathfinding’. He will be building on his recent work developing Lazy CBS, which applies nogood learning techniques to multi-agent pathfinding, to design scalable, robust pathfinding algorithms. Congratulations, Graeme!
Page 1 / 3 »