On-chip probabilistic inference for charged-particle tracking at the sensor edge

TL;DR

It is demonstrated that neural networks embedded in the front-end electronics can infer charged-particle kinematic parameters from a single silicon layer, establishing a path toward probabilistic inference directly at the edge, opening new opportunities for intelligent sensing in high-rate scientific instruments.

Abstract

Modern scientific instruments operate under increasingly extreme constraints on bandwidth, latency, and power. Inference at the sensor edge determines experimental data collection efficiency by deciding which information to save for further analysis. Particle tracking detectors at the Large Hadron Collider exemplify this challenge: pixelated silicon sensors generate rich spatiotemporal ionization patterns, yet most of this information is discarded due to data-rate limitations. Concurrently, advancements in co-design tools provide rapid turn-around for incorporating machine learning into application-specific integrated circuits, motivating designs for particle detectors with new integrated technologies. We demonstrate that neural networks embedded in the front-end electronics can infer charged-particle kinematic parameters from a single silicon layer. We regress hit positions and incident angles with calibrated uncertainties, while satisfying stringent constraints on numerical precision, latency, and silicon area. Our results establish a path toward probabilistic inference directly at the edge, opening new opportunities for intelligent sensing in high-rate scientific instruments.

Figures (8)

• Figure 1: The silicon sensor is described by a Cartesian coordinate system where $x$ is in the direction of LHC beams and $y$ is normal to the $x$-axis. The $z$-direction, used to define the angles $\alpha$ and $\beta$, is perpendicular through the depletion region of the sensor pointing outwards from the interaction region of the experiment.
• Figure 2: Top: Threshold values as a function of epoch in the training of the Max transformer with SoftQuantize. Bottom: Points indicate the preferred charge thresholds for each regression model. The vertical lines and shaded bands represent, respectively, the means and standard deviations of the thresholds obtained by the transformer.
• Figure 3: Comparison of the residuals for $x$, $y$, $\alpha$, and $\beta$ for all models. The solid lines represent the minimum interval containing 68% of clusters. For the Max and Full models, the shaded bands represent the mean predicted uncertainty on each variable. Across architectures, the dominant performance degradation arises from temporal sparsification of the input, whereas aggressive charge quantization introduces comparatively minor loss.
• Figure 4: Comparison of the residuals for $x$, $y$, $\alpha$, and $\beta$ between the Full MLP model and non-ML algorithms. The non-ML algorithms labeled as "optimistic" incorporate information that is not directly available on the ASIC.
• Figure : (a)