GiPH: Generalizable Placement Learning for Adaptive Heterogeneous Computing

Careful placement of a computational application within a target device cluster is critical for achieving low application completion time. The problem is challenging due to its NP-hardness and combinatorial nature. In recent years, learning-based approaches have been proposed to learn a placement policy that can be applied to unseen applications, motivated by the problem of placing a neural network across cloud servers. These approaches, however, generally assume the device cluster is fixed, which is not the case in mobile or edge computing settings, where heterogeneous devices move in and out of range for a particular application. We propose a new learning approach called GiPH, which learns policies that generalize to dynamic device clusters via 1) a novel graph representation gpNet that efficiently encodes the information needed for choosing a good placement, and 2) a scalable graph neural network (GNN) that learns a summary of the gpNet information. GiPH turns the placement problem into that of finding a sequence of placement improvements, learning a policy for selecting this sequence that scales to problems of arbitrary size. We evaluate GiPH with a wide range of task graphs and device clusters and show that our learned policy rapidly find good placements for new problem instances. GiPH finds placements with up to 30.5% lower completion times, searching up to 3X faster than other search-based placement policies.

Accurate Cooperative Sensor Fusion by Parameterized Covariance Generation for Sensing and Localization Pipelines in CAVs

A major challenge in cooperative sensing is to weight the measurements taken from the various sources to get an accurate result. Ideally, the weights should be inversely proportional to the error in the sensing information. However, previous cooperative sensor fusion approaches for autonomous vehicles use a fixed error model, in which the covariance of a sensor and its recognizer pipeline is just the mean of the measured covariance for all sensing scenarios. The approach proposed in this paper estimates error using key predictor terms that have high correlation with sensing and localization accuracy for accurate covariance estimation of each sensor observation. We adopt a tiered fusion model consisting of local and global sensor fusion steps. At the local fusion level, we add in a covariance generation stage using the error model for each sensor and the measured distance to generate the expected covariance matrix for each observation. At the global sensor fusion stage we add an additional stage to generate the localization covariance matrix from the key predictor term velocity and combines that with the covariance generated from the local fusion for accurate cooperative sensing. To showcase our method, we built a set of 1/10 scale model autonomous vehicles with scale accurate sensing capabilities and classified the error characteristics against a motion capture system. Results show an average and max improvement in RMSE when detecting vehicle positions of 1.42x and 1.78x respectively in a four-vehicle cooperative fusion scenario when using our error model versus a typical fixed error model.