Search for: All records

Path planning is a critical task for autonomous driving, aiming to generate smooth, collision-free, and feasible paths based on input perception and localization information. The planning task is both highly time-sensitive and computationally intensive, posing significant challenges to resource-constrained autonomous driving hardware. In this article, we propose an end-to-end framework for accelerating path planning on FPGA platforms. This framework focuses on accelerating quadratic programming (QP) solving, which is the core of optimization-based path planning and has the most computationally-intensive workloads. Our method leverages a hardware-friendly alternating direction method of multipliers (ADMM) to solve QP problems while employing a highly parallelizable preconditioned conjugate gradient (PCG) method for solving the associated linear systems. We analyze the sparse patterns of matrix operations in QP and design customized storage schemes along with efficient sparse matrix multiplication and sparse matrix-vector multiplication units. Our customized design significantly reduces resource consumption for data storage and computation while dramatically speeding up matrix operations. Additionally, we propose a multi-level dataflow optimization strategy. Within individual operators, we achieve acceleration through parallelization and pipelining. For different operators in an algorithm, we analyze inter-operator data dependencies to enable fine-grained pipelining. At the system level, we map different steps of the planning process to the CPU and FPGA and pipeline these steps to enhance end-to-end throughput. We implement and validate our design on the AMD ZCU102 platform. Our implementation achieves state-of-the-art performance in both latency and energy efficiency compared with existing works, including an average 1.48× speedup over the best FPGA-based design, a 2.89× speedup compared with the state-of-the-art QP solver on an Intel i7-11800H CPU, a 5.62× speedup over an ARM Cortex-A57 embedded CPU, and a 1.56× speedup over state-of-the-art GPU-based work. Furthermore, our design delivers a 2.05× improvement in throughput compared with the state-of-the-art FPGA-based design. 

Abstract. Highly oxygenated organic molecules (HOMs) from α-pinene ozonolysis have been shown to be significant contributors to secondary organic aerosol (SOA), yet our mechanistic understanding of how the peroxy-radical-driven autoxidation leads to their formation in this system is still limited. The involved isomerisation reactions such as H-atom abstractions followed by O2 additions can take place on sub-second timescales in short-lived intermediates, making the process challenging to study. Similarly, while the end-products and sometimes radical intermediates can be observed using mass spectrometry, their structures remain elusive. Therefore, we propose a method utilising selective deuterations for unveiling the mechanisms of autoxidation, where the HOM products can be used to infer which C atoms have taken part in the isomerisation reactions. This relies on the fact that if a C−D bond is broken due to an abstraction by a peroxy group forming a −OOD hydroperoxide, the D atom will become labile and able to be exchanged with a hydrogen atom in water vapour (H2O), effectively leading to loss of the D atom from the molecule. In this study, we test the applicability of this method using three differently deuterated versions of α-pinene with the newly developed chemical ionisation Orbitrap (CI-Orbitrap) mass spectrometer to inspect the oxidation products. The high mass-resolving power of the Orbitrap is critical, as it allows the unambiguous separation of molecules with a D atom (mD=2.0141) from those with two H atoms (mH2=2.0157). We found that the method worked well, and we could deduce that two of the three tested compounds had lost D atoms during oxidation, suggesting that those deuterated positions were actively involved in the autoxidation process. Surprisingly, the deuterations were not observed to decrease HOM molar yields, as would have been expected due to kinetic isotope effects. This may be an indication that the relevant H (or D) abstractions were fast enough that no competing pathways were of relevance despite slower abstraction rates of the D atom. We show that selective deuteration can be a very useful method for studying autoxidation on a molecular level and likely is not limited to the system of α-pinene ozonolysis tested here. 

Abstract. Oxygenated organic molecules (OOMs) are the crucial intermediates linkingvolatile organic compounds (VOCs) to secondary organic aerosols (SOAs) in theatmosphere, but comprehensive understanding of the characteristics of OOMsand their formation from VOCs is still missing. Ambient observations ofOOMs using recently developed mass spectrometry techniques are stilllimited, especially in polluted urban atmospheres where VOCs and oxidants areextremely variable and complex. Here, we investigate OOMs, measured by anitrate-ion-based chemical ionization mass spectrometer at Nanjing ineastern China, through performing positive matrix factorization on binnedmass spectra (binPMF). The binPMF analysis reveals three factors aboutanthropogenic VOC (AVOC) daytime chemistry, three isoprene-relatedfactors, three factors about biogenic VOC (BVOC) nighttime chemistry, andthree factors about nitrated phenols. All factors are influenced by NOxin different ways and to different extents. Over 1000 non-nitro moleculeshave been identified and then reconstructed from the selected solution ofbinPMF, and about 72 % of the total signals are contributed bynitrogen-containing OOMs, mostly regarded as organic nitrates formed throughperoxy radicals terminated by nitric oxide or nitrate-radical-initiatedoxidations. Moreover, multi-nitrates account for about 24 % of the totalsignals, indicating the significant presence of multiple generations,especially for isoprene (e.g., C5H10O8N2 andC5H9O10N3). Additionally, the distribution of OOMconcentration on the carbon number confirms their precursors are driven by AVOCsmixed with enhanced BVOCs during summer. Our results highlight the decisiverole of NOx in OOM formation in densely populated areas, and we encouragemore studies on the dramatic interactions between anthropogenic and biogenicemissions.