WG211/M15Nardi

Title: Vertically-integrated exploration of algorithmic and implementation design spaces in 3D scene understanding

Speaker: Luigi Nardi ([1])

Abstract: Real-time computer vision and in particular simultaneous localisation and mapping (SLAM) offer great potential for a new level of scene modelling, tracking and real environmental interaction for many types of robot, but their high computational requirements mean that use on mass market embedded platforms is challenging. Meanwhile, trends in low-cost, low-power processing are towards massive parallelism and heterogeneity, making it difficult for robotics and computer vision researchers to implement their algorithms in a performance-portable way.

In this talk we briefly introduce SLAMBench, a publicly-available benchmarking framework which represents a starting point for quantitative, comparable and validatable experimental research to investigate trade-offs in performance, accuracy and energy consumption of a SLAM system. We then examine how it can be mapped to power constrained embedded systems. Key to our approach is the idea of incremental co-design exploration, where optimisation choices that concern the domain layer are incrementally explored together with low-level compiler and architecture choices. The goal of this exploration is to reduce execution time while minimising the power but without sacrificing the quality of the result. We can see a clear interest in pushing the envelope at a higher level of abstraction than current popular languages, e.g. C. The take-away message is that domain-specific knowledge can give a boost in performance, a boost that would not have been harnessed by more standard compiler optimisation techniques.