I recently attended to the 2012 International Conference on Automated Planning and Scheduling (ICAPS), which was held at a car distance from São Paulo (what a rare opportunity!). It was the first time that I saw something like Festivus, which featured funny debates about the relevance of the research developed by the P&S community and even a bossa nova song about three blocks that wanted to be moved. Beyond the fun and cheap transportation cost, the most remarking experiences I had there were the satellite event to gather and mentor graduate students – ICAPS Doctoral Consortium – as well as observing the differences between planners and schedulers – the former of which somewhat belonging to the OR community.
ICAPS Doctoral Consortium (DC)
DC consisted of four talks directed to PhD students and a poster presentation, during which at least two mentors were assigned to talk with each student to discuss his/her research topic.
Alan Fern managed to invite young and mid-career researchers that followed varied directions after earning their PhD and were willing to tell their stories to us.
Andrew Coles presented a talk entitle “Now what?” describing his career and how much time does it take to achieve certain positions in the academia.
Silvia Richter’s talk was mostly focused on issues such as finding motivation and realizing that it takes a lot of time to have a good idea. She also provided good figures regarding how many papers you should publish (3-4 is nice) and how much effort should you put on writing your thesis (not as much as you suppose, since only 5 people will probably read it entirely), and suggested an interesting website: 3monththesis.com.
Minh Do showed some interesting graphs to compare salaries, freedom, possibilities for changing career afterwards etc for positions like working in the industry, researching in the industry, researching in the academia and striving to be a professor with tenure. On top of that, he listed four activities that any CS undergrad student should experience to know what to do with his/her life afterwards (and that I wish I was told years ago):
- Do a research internship in academia
- Do an internship at the headquarters of a company
- Work in a local startup with a good tech team
- Work in an open source project
Scott Sanner finished the session with good tips for social networking and external presence. In the former case, he stressed the importance of talking to other researchers and getting to know more about their work (and not bragging about yours) and of giving memorable talks, or at least striving to do that. In the latter case, he talked about building a professional website and he showed his first website as a funny example of what you should not do.
I presented a poster about adaptive search methods for Constraint-Based Scheduling (CBS). The first mentor who showed up was Amanda Coles (wife of Andrew Coles, the first speaker above). We had a long and interesting talk, that let me realize how much of the theoretical background of adaptive search methods for combinatorial optimization is also valuable for planning. The second mentor was Stephen Smith, who was the advisor of a number of works related to my topic of interest. I was glad that both mentors enjoyed my research plan and gave me good advices to succeed with it.
Planning vs. Scheduling (and OR)
I have already mentioned in a former post the divide between planners and schedulers: the summer school consisted of three courses about planning and one about scheduling, the latter being easier to me than the others. While some recognize that planning methods can theoretically be used to solve scheduling problems, planners strive to be as generalist as possible when approaching a problem. Schedulers, however, prefer the opposite path and delve themselves into the structure of each problem to take the most of it. In practice, their application domains barely touch each other. Nevertheless, there are some researchers that, as we say in Brazil, are on the top of the wall that divides those areas. Even though I was feeling an outsider at some moments, there were a number of talks related to OR and a few others discussing the limits of application of each type of technique. Thus, tying together planners and schedulers seems to be a good long-term strategy to both areas.