I had so much fun with The Application Game that I decided to play again googling “constraint satisfaction problem” and “refrigerator”. I came up with “Service robot planning via solving constraint satisfaction problem“.

The problem of demographic shifts towards the elderly is deteriorating, as the relative number of caregivers is insufficient to provide the support required for their wellbeing, which is further aggravated by the increasingly hectic lifestyle. Service robot is getting more prominent as a possible solution. Robot manipulation and mobility is an important field, but they also require high level planning for these minute actions in order to provide ample support. Automatic service composition, contributed significantly by web services, offers the necessary technology for the task. Robot planning problem can be solved by representing it as constraint satisfaction problem (CSP) due to it being able to support loose binding of services and variables of wider domain. 

What does this have to do with refrigerators?

Given a certain goal, planner module will compose a sequence of plans for the robot to execute such that after the plans are completed, the goal is achieved. For example, given that the goal is to place a canned soft drink in the fridge and that all cabinet needs to be closed, the robot will first find where the soft drink is. If there is one that happens to be in the kitchen cabinet, the robot will approach it, open the cabinet and pick up the soft drink, and then close it. It will subsequently approach and open the fridge, after which it will place the canned drink before closing it. No prior programming is required for the plan.

Ok, tenuous, but this paper illustrates an interesting point.

The paper is published in the ROBOMECH Journal. Its authors are in a Graduate School of System Design, not a computer science department. The closest it comes to referencing the CP literature seems to be a 2008 paper by De Moura and BjØrner on Z3: An efficient SMT solver, in the International Conference on Tools and Algorithms for the Construction and Analysis of Systems and a 2014 paper by Georgievski and Aiello, “An Overview of Hierarchical Task Network Planning” in ArXiv.

In other words, your typical attendee at the CP conference or reader of the Constraints journal might well not have run across this paper. I believe there are many such papers. In a way, this is a positive sign, of the degree to which our “baby” has moved out into the world. Still, the more we are aware of applications of constraint satisfaction the better. The more that people applying constraint satisfaction are aware of the broader body of work in our field the better. The more that people in general are aware of the breadth of applications of constraint satisfaction the better. An application like this one to service robots could help interest students in the field or convince funding agencies of our “impact”. It could suggest new collaboration opportunities or suggest new directions for basic research.