Short-term prediction of construction waste transport activities using AI-Truck

TL;DR

The paper tackles short-term, city-scale prediction of slag-truck stay-point concentrations during heavy pollution episodes to aid environmental enforcement. It introduces AI-Truck, a deep ensemble framework that fuses Bi-LSTM, TCN, STGCN, and PDFormer through a soft voting mechanism, and addresses spatial data imbalance via downsampling and weighted loss. A key contribution is the OD-based geographic feature design combined with semantic similarity (FastDTW) and temporal patterns to capture complex spatio-temporal dependencies. Experiments in Chengdu demonstrate superior macro F1 ($=0.747$) and practical foresight (high-activity areas identifiable up to $1.5$ hours ahead with $>80\%$ accuracy), underscoring the method’s value for proactive environmental management.

Abstract

Construction waste hauling trucks (or `slag trucks') are among the most commonly seen heavy-duty diesel vehicles in urban streets, which not only produce significant carbon, NO$_{\textbf{x}}$ and PM$_{\textbf{2.5}}$ emissions but are also a major source of on-road and on-site fugitive dust. Slag trucks are subject to a series of spatial and temporal access restrictions by local traffic and environmental policies. This paper addresses the practical problem of predicting levels of slag truck activity at a city scale during heavy pollution episodes, such that environmental law enforcement units can take timely and proactive measures against localized truck aggregation. A deep ensemble learning framework (coined AI-Truck) is designed, which employs a soft vote integrator that utilizes Bi-LSTM, TCN, STGCN, and PDFormer as base classifiers. AI-Truck employs a combination of downsampling and weighted loss is employed to address sample imbalance, and utilizes truck trajectories to extract more accurate and effective geographic features. The framework was deployed for truck activity prediction at a resolution of 1km$\times$1km$\times$0.5h, in a 255 km$^{\textbf{2}}$ area in Chengdu, China. As a classifier, AI-Truck achieves a macro F1 of 0.747 in predicting levels of slag truck activity for 0.5-h prediction time length, and enables personnel to spot high-activity locations 1.5 hrs ahead with over 80\% accuracy.

Figures (15)

• Figure 1: Regional categories of stay points.
• Figure 2: Spatial distribution of $v_s^{t}$ in Chengdu's KMAs.
• Figure 3: Cumulative distribution function of $v_s^a$ during heavy pollution episodes.
• Figure 4: Spatial distribution of $v_s^{t}$ after down-sampling in Chengdu's KMAs.
• Figure 5: Cumulative distribution function of $v_s^t$ during heavy pollution episodes.