Smart techniques for flying-probe testing

In the production of printed circuit boards, in-circuit tests verify whether the electric and electronic components of the board have been correctly soldered. When the test is performed using flying-probes, several probes are simultaneously moved on the board to reach and touch multiple test points. Taking into consideration the layout of the board, the characteristics of the tester, and several other physical constraints, not all movements of the probes are mutually compatible nor they can always be performed through simple straight lines. As the cost of the test is mainly related to its length, and patching the path of one probe may create new incompatibilities with the trajectory of the other probes, one should carefully trade off the time required to find the trajectories with the time required by the probes to follow them. In this paper, we model the movements of our flying probes as a multiple and collaborative planning problem. We describe an approach for detecting invalid movements and we design a strategy to correct them with the addition of new intermediate points in the trajectory. We report the entire high-level procedure and we explore the optimizations performed in the more expensive and complex steps. We also present parallel implementations of our algorithms, either relying on multi-core CPU devices or many-cores GPU platforms, when these units may be useful to achieve greater speedups. Experimental results show the effectiveness of the proposed solution in terms of elapsed computation times.