Occupancy Prediction Research Articles

Mutual information-driven self-supervised point cloud pre-training

Learning universal representations from unlabeled 3D point clouds is essential to improve the generalization and safety of autonomous driving. Generative self-supervised point cloud pre-training with low-level features as pretext tasks is a mainstream paradigm. However, from the perspective of mutual information, this approach is constrained by spatial information and entangled representations. In this study, we propose a generalized generative self-supervised point cloud pre-training framework called GPICTURE. High-level features were used as an additional pretext task to enhance the understanding of semantic information. Considering the varying difficulties caused by the discrimination of voxel features, we designed inter-class and intra-class discrimination-guided masking (I2Mask) to set the masking ratio adaptively. Furthermore, to ensure a hierarchical and stable reconstruction process, centered kernel alignment-guided hierarchical reconstruction and differential-gated progressive learning were employed to control multiple reconstruction tasks. Complete theoretical analyses demonstrated that high-level features can enhance the mutual information between latent features and high-level features, as well as the input point cloud. On Waymo, nuScenes, and SemanticKITTI, we achieved a 75.55% mAP for 3D object detection, 79.7% mIoU for 3D semantic segmentation, and 18.8% mIoU for occupancy prediction. Specifically, with only 50% of the fine-tuning data required, the performance of GPICURE was close to that of training from scratch with 100% of the fine-tuning data. In addition, consistent visualization with downstream tasks and a 57% reduction in weight disparity demonstrated a better fine-tuning starting point. The project page is hosted at https://gpicture-page.github.io/.

RaNAS: Resource-Aware Neural Architecture Search for Edge Computing

Neural architecture search (NAS) for edge devices is often time-consuming because of long-latency deploying and testing on edge devices. The ability to accurately predict the computation cost and memory requirement for convolutional neural networks (CNNs) in advance holds substantial value. Existing work primarily relies on analytical models, which can result in high prediction errors. This paper proposes a resource-aware NAS (RaNAS) model based on various features. Additionally, a new graph neural network is introduced to predict inference latency and maximum memory requirements for CNNs on edge devices. Experimental results show that, within the error bound of ±1%, RaNAS achieves an accuracy improvement of approximately 8% for inference latency prediction and about 25% for maximum memory occupancy prediction over the state-of-the-art approaches.

TDOcc: Exploit machine learning and big data in multi-view 3D occupancy prediction

With the advancement of machine learning and big data technologies, BEV(Bird’s Eye View)-based methodologies have recently achieved significant breakthroughs in multi-view 3D occupancy prediction tasks. However, BEV(Bird’s Eye View)-centric 3D occupancy prediction continues to grapple with feature representation and annotation costs when applied to complex open environments. In order to surmount these issues and further propel the evolution of 3D occupancy tasks, this study introduces a novel framework termed TDOcc. By leveraging multi-camera imagery, TDOcc executes 3D semantic occupancy prediction by directly learning from unprocessed 3D spaces, thereby maximizing information retention. TDOcc presents two notable advantages: firstly, it utilizes dense occupancy labels, which not only facilitate robust dense occupancy inference but also enable comprehensive object estimation within the scene. Secondly, the framework synthesizes historical feature information by adeptly aligning past and present features through temporal cues, thereby bolstering the efficacy of the feature fusion module. Additionally, with a view to address the ill-posed nature inherent in camera-based 3D occupancy prediction, we incorporate an enhancement module that operates within the 3D feature space. This module has been meticulously crafted for the training phase to amplify the model’s learning potential. Extensive experiments conducted on the widely recognized nuScenes dataset underscore the efficacy of our proposed approach. Compared to the most recent TPVFormer and OccFormer, our approach has achieved a significant improvement in mean Intersection over Union (mIoU) by 2.0 and 0.8 respectively, and has reached performance comparable to the state-of-the-art LiDAR-based methods.

A Sparse Octree-Based CNN for Probabilistic Occupancy Prediction Applied to Next Best View Planning

A Sparse Octree-Based CNN for Probabilistic Occupancy Prediction Applied to Next Best View Planning

Broad-scale predictions of herpetofauna occupancy and colonization in an agriculturally dominated landscape.

Predictions of species occurrence allow land managers to focus conservation efforts on locations where species are most likely to occur. Such analyses are rare for herpetofauna compared to other taxa, despite increasing evidence that herptile populations are declining because of landcover change and habitat fragmentation. Our objective was to create predictions of occupancy and colonization probabilities for 15 herptiles of greatest conservation need in Iowa. From 2006-2014, we surveyed 295 properties throughout Iowa for herptile presence using timed visual-encounter surveys, coverboards, and aquatic traps. Data were analyzed using robust design occupancy modeling with landscape-level covariates. Occupancy ranged from 0.01 (95% CI = -0.01, 0.03) for prairie ringneck snake (Diadophis punctatus arnyi) to 0.90 (95% CI = 0.898, 0.904) for northern leopard frog (Lithobates pipiens). Occupancy for most species correlated to landscape features at the 1-km scale. General patterns of species' occupancy included negative effects of agricultural features and positive effects of water features on turtles and frogs. Colonization probabilities ranged from 0.007 (95% CI = 0.006, 0.008) for spiny softshell turtle (Apalone spinifera) to 0.82 (95% CI = 0.62, 1.0) for western fox snake (Pantherophis ramspotti). Colonization probabilities for most species were best explained by effects of water and grassland landscape features. Predictive models had strong support (AUC > 0.70) for six out of 15 species (40%), including all three turtles studied. Our results provide estimates of occupancy and colonization probabilities and spatial predictions of occurrence for herptiles of greatest conservation need across the state of Iowa.

Spectrum occupancy prediction using LSTM models for cognitive radio applications

ABSTRACT In recent days, mobile traffic prediction has become a prominent solution for spectrum management-related operations for the next-generation cellular networks in Cognitive Radio (CR) applications. To achieve this, the binary dataset has been created from the captured data by monitoring the spectrum activities of nine different Long Term Evolution (LTE) frequency channels. We propose a Long Short Term Memory (LSTM) based Spectrum Occupancy Prediction (SOP) approach for modelling infrastructure-based cellular traffic systems. The different types of LSTM models, such as Convolutional, Convolutional Neural Network (CNN), Stacked, and Bidirectional have been generated via offline training and tested for the created binary datasets. Moreover, the prediction performance evaluation of the generated LSTM models has been calculated using Mean Absolute Error (MAE). The pro- posed LSTM-based SOP model has achieved 2.5% higher prediction accuracy than the Auto-Regressive Integrated Moving Average (ARIMA) statistical model, accurately aligning the traffic trend with the actual samples.

Self-Supervised 3D Semantic Occupancy Prediction from Multi-View 2D Surround Images

An accurate 3D representation of the geometry and semantics of an environment builds the basis for a large variety of downstream tasks and is essential for autonomous driving related tasks such as path planning and obstacle avoidance. The focus of this work is put on 3D semantic occupancy prediction, i.e., the reconstruction of a scene as a voxel grid where each voxel is assigned both an occupancy and a semantic label. We present a Convolutional Neural Network-based method that utilizes multiple color images from a surround-view setup with minimal overlap, together with the associated interior and exterior camera parameters as input, to reconstruct the observed environment as a 3D semantic occupancy map. To account for the ill-posed nature of reconstructing a 3D representation from monocular 2D images, the image information is integrated over time: Under the assumption that the camera setup is moving, images from consecutive time steps are used to form a multi-view stereo setup. In exhaustive experiments, we investigate the challenges presented by dynamic objects and the possibilities of training the proposed method with either 3D or 2D reference data. Latter being motivated by the comparably higher costs of generating and annotating 3D ground truth data. Moreover, we present and investigate a novel self-supervised training scheme that does not require any geometric reference data, but only relies on sparse semantic ground truth. An evaluation on the Occ3D dataset, including a comparison against current state-of-the-art self-supervised methods from the literature, demonstrates the potential of our self-supervised variant.

The prediction of thermal sensation in building using support vector machine and extreme gradient boosting

The building has great potential for energy savings as one of the locations that humans often occupy. In addition to energy efficiency, humans must consider environmental sustainability and the comfort of the building's occupants. Conditioning of indoor air quality, including those related to thermal comfort, continues to be pursued to be more economical, one of which is to utilize the prediction of occupants' thermal sensations. The prediction results can be utilized to adjust room air conditions more economically. This paper proposes using extreme gradient boosting (XGBoost) and support vector machine (SVM) to predict the thermal sensation in the building. The built environment parameters are preprocessed, and the thermal sensation is predicted by intelligent systems. The ten variables that most influence the level of accuracy of this thermal sensation prediction system are thermal preference vote, indoor operative temperature, Griffith's neutral temperature, indoor globe temperature, mean radiant temperature, Indoor air temperature, predicted mean vote, and outdoor mean temperature. SVM with four features, XGBoost and XGBoost with hyperparameter tuning, achieve an accuracy of 99.45%, 97.81%, and 98.08%, respectively. Regarding computational complexity, training an SVM system with the same number of features requires a shorter time than XGBoost training. The same thing also happened with the test of the SVM system, which required a shorter time compared to the time for the examination of the XGBoost system.

HybridOcc: NeRF Enhanced Transformer-Based Multi-Camera 3D Occupancy Prediction

HybridOcc: NeRF Enhanced Transformer-Based Multi-Camera 3D Occupancy Prediction

CRISP-DM-Based Data-Driven Approach for Building Energy Prediction Utilizing Indoor and Environmental Factors

The significant energy consumption associated with the built environment demands comprehensive energy prediction modelling. Leveraging their ability to capture intricate patterns without extensive domain knowledge, supervised data-driven approaches present a marked advantage in adaptability over traditional physical-based building energy models. This study employs various machine learning models to predict energy consumption for an office building in Berkeley, California. To enhance the accuracy of these predictions, different feature selection techniques, including principal component analysis (PCA), decision tree regression (DTR), and Pearson correlation analysis, were adopted to identify key attributes of energy consumption and address collinearity. The analyses yielded nine influential attributes: heating, ventilation, and air conditioning (HVAC) system operating parameters, indoor and outdoor environmental parameters, and occupancy. To overcome missing occupancy data in the datasets, we investigated the possibility of occupancy-based Wi-Fi prediction using different machine learning algorithms. The results of the occupancy prediction modelling indicate that Wi-Fi can be used with acceptable accuracy in predicting occupancy count, which can be leveraged to analyze occupant comfort and enhance the accuracy of building energy models. Six machine learning models were tested for energy prediction using two different datasets: one before and one after occupancy prediction. Using a 10-fold cross-validation with an 8:2 training-to-testing ratio, the Random Forest algorithm emerged superior, exhibiting the highest R2 value of 0.92 and the lowest RMSE of 3.78 when occupancy data were included. Additionally, an error propagation analysis was conducted to assess the impact of the occupancy-based Wi-Fi prediction model’s error on the energy prediction model. The results indicated that Wi-Fi-based occupancy prediction can improve the data inputs for building energy models, leading to more accurate energy consumption predictions. The findings underscore the potential of integrating the developed energy prediction models with fault detection systems, model predictive controllers, and energy load shape analysis, ultimately enhancing energy management practices.

Intelligent detection of office occupancy using hybrid data-mining

The importance of office occupancy detection is evident, due to its critical role in workplace management and facility optimization. Recent studies have recommended the use of data-mining techniques for this purpose. In this research, a powerful hybrid model is developed upon artificial neural network (ANN) coupled with shuffled complex evolution (SCE) algorithm for the binary prediction of occupancy in an office room based on several characteristics such as light, temperature, humidity, and CO2. The trained model is tested via two datasets (considering the opened/closed office door while occupied) and according to the results, it achieved an excellent prediction. In detail, the values of area under the curve (AUC) were above 0.98 along with mean square errors below 0.03; indicating the superiority of the proposed SCE algorithm compared to three baseline techniques, namely Archimedes optimization algorithm (AOA), incomprehensible but Intelligible-in-time logics (ILA), and gazelle optimization algorithm (GOA) in this study, as well as some others in earlier literature. Moreover, the SCE-ANN emerged as the most efficient (i.e., the quickest and least complex) algorithm. In the end, the dataset is exposed to a well-known statistical method called principal component analysis (PCA) to evaluate the importance of the considered input factors and identify the most crucial ones. Altogether, the findings of this study can provide valuable environmental and economic contributions to related sectors such as building energy management.

Research on occupant injury severity prediction of autonomous vehicles based on transfer learning

AbstractThe focus of the future of autonomous vehicles has shifted from feasibility to safety and comfort. The seat of an autonomous vehicle may be equipped with a rotational function, and the occupant's sitting position would be diverse. This poses a higher challenge to occupant injury protection during vehicle collisions. The main objective of the current study is to develop occupant injury prediction models for autonomous vehicles that can be used to predict the injury severity of occupants in different seat orientations and sitting positions. The first step is to establish an occupant crash model database with different seat orientations. It is used to simulate the occupant crash injury database of an autonomous vehicle, considering seat rotation and the back inclination angle. The second step is to establish a pre‐training occupant injury prediction model based on the existing database and then train the autonomous vehicle occupant injury prediction model using an in‐house database based on the transfer learning method. Occupant injury prediction models achieve good accuracy (82.8% on the numerical database and 62.9% on the real verification database) and shorter computational time (4.86 ± 0.33 ms) on the prediction tasks. Finally, the influence of the model input variables is analyzed. This study demonstrates the feasibility of using a small‐sample database based on transfer learning for occupant injury prediction in autonomous vehicles.

Prediction of residential building occupancy using Machine learning with integrated sensor and survey Data: Insights from a living lab in Morocco

Building occupancy information is essential for effective energy management in buildings through the adoption of energy conservation and occupant-centric control strategies. These strategies endeavor to contribute to optimizing energy consumption while ensuring occupant comfort. This study focuses on advanced occupancy modeling techniques to enhance energy efficiency in residential buildings, utilizing various data-driven techniques. Various Machine Learning models, such as Random Forest, Bayesian Network, Decision Trees, Support Vector Machines, K-Nearest Neighbors, eXtreme Gradient Boosting, and Regularized Greedy Forest, have been evaluated for predicting and classifying building occupancy. Employing a living lab approach within a residential setting, the research evaluates model performance using two ground truth datasets: IoT sensor and survey data. Experimental tests use Accuracy and F1_score as evaluation criteria, demonstrating accuracy rates ranging from 70% to 95.96%. The results highlight the performance of these models in predicting residential occupancy, offering valuable insight into two approaches to occupancy modeling. The Random Forest model performs exceptionally well in capturing occupancy trends, while the Bayesian Network model, combined with expert knowledge, provides detailed predictions of occupancy types and zones. The research also addresses challenges related to data collection, including privacy concerns. It presents effective occupancy modeling strategies for residential energy management.

Advancing smart tourism destinations: A case study using bidirectional encoder representations from transformers‐based occupancy predictions in torrevieja (Spain)

AbstractTourism represents a crucial socio‐economic pillar globally, yet the multifaceted challenges it poses necessitate innovative management approaches. The paradigm of smart tourism harnesses advanced data analytics tools to promote both profitability and sustainability in tourist destinations, leading to new levels of destination smartness. Accurate tourist occupancy prediction, particularly in areas dominated by second‐home accommodations where traditional hospitality data may be insufficient, plays a key role in optimising tourism management. To address this data gap, our prior research employed ARIMA modelling on Airbnb booking time series and analysed tourism‐related Twitter conversations to forecast occupancy levels in Torrevieja (Alicante); a prominent second‐home tourism destination in Southeastern Spain. In this extended study, we delve deeper into the realm of social sensing by utilising bidirectional encoder representations from transformers (BERT) for topic modelling. Our methodology involves the processing and analysis of Twitter data to identify prominent themes related to Torrevieja. The findings not only reveal nuanced perceptions and discussions about the destination but also underscore the effectiveness of BERT in capturing intricate topic dynamics. Importantly, this work highlights how the alignment of specific topics with booking patterns can further enhance predictive accuracy for tourist occupancy, presenting a robust toolkit for stakeholders in the tourism sector.

건물 에너지의 효율적 관리를 위한 저해상도 열화상 이미지를 활용한 재실자 수 예측에 관한 연구

건물 에너지의 효율적 관리를 위한 저해상도 열화상 이미지를 활용한 재실자 수 예측에 관한 연구

Enhanced multi-horizon occupancy prediction in smart buildings using cascaded Bi-LSTM models with integrated features

Accurate occupancy prediction in smart buildings is crucial for optimizing energy management, improving occupant comfort, and effectively controlling building systems, particularly for short- and long-term horizons. Recently, deep learning-based occupancy prediction methods have gained considerable attention. However, the full potential of these methods remains under explored in terms of model architecture variations and prediction horizons. This study introduces cascaded LSTM and cascaded Bi-LSTM models for multi-horizon predictions from 10 minutes to 24 hours, integrating a modified activation function, additional input features, and optimized hyper-parameters using OPTUNA. Traditional performance metrics and various other analyses were conducted to compare the models. Both models performed well for short- and long-term predictions, with minimal differences in the results. Nevertheless, analysis focusing on non-zero data errors (accounting for approximately 11% of occupied periods) and occupancy-wise errors showed a significant performance gap between the two models. The cascaded Bi-LSTM model demonstrated consistent performance across various prediction horizons and occupancy variations, with accuracy approximately 10-15% higher than the cascaded LSTM model, highlighting its superior capability in capturing complex dataset dynamics through a bidirectional process. This study highlights the importance of additional input features, data feature analysis, and multi-perspective result analysis to select the most suitable model for occupancy prediction, validated with pre- and post-modeling feature importance analysis.

Enabling active visitor management: local, short-term occupancy prediction at a touristic point of interest

After the temporary shock of the Covid-19 pandemic, the rapid recovery and resumed growth of the tourism sectors accelerates unsustainable tourism, resulting in local (over-)crowding, environmental damage, increased emissions, and diminished tourism acceptance. Addressing these challenges requires an active visitor management system at points of interest (POI), which requires local and timely POI-specific occupancy predictions to predict and mitigate crowding. Therefore, we present a new approach to measure visitor movement at an open-spaced, and freely accessible POI and evaluate the prediction performance of multiple occupancy and visitor count machine learning prediction models. We analyze multiple case combinations regarding spatial granularity, time granularity, and prediction time horizons. With an analysis of the SHAP values we determine the influence of the most important features on the prediction and extract transferable knowledge for similar regions lacking visitor movement data. The results underline that POI-specific prediction is achievable with a moderate relation for occupancy prediction and a strong relation for visitor count prediction. Across all cases, XGBoost and Random Forest outperform other models, with prediction accuracy increasing as the prediction time horizon shortens. For effective active visitor management, combining multiple models with different spatial aggregations and prediction time horizons provides the best information basis to identify appropriate steering measures. This innovative application of digital technologies facilitates information exchange between destination management organizations and tourists, promoting sustainable destination development and enhancing tourism experience.

Occupation Prediction with Multimodal Learning from Tweet Messages and Google Street View Images

Occupation Prediction with Multimodal Learning from Tweet Messages and Google Street View Images

Unraveling the characteristic spatial scale of habitat selection for forest grouse species in the boreal landscape

The characteristic spatial scale at which species respond strongest to forest structure is unclear and species-specific and depends on the degree of landscape heterogeneity. Research often analyzes a pre-defined spatial scale when constructing species distribution models relating forest variables with occupancy patterns. This is a limitation, as forest characteristics shape the species use of habitat at multiple spatial scales. To explore the drivers of this relationship, we conducted an in-depth investigation into how scaling forest variables at biologically relevant spatial scales affects occupancy of grouse species in boreal forest. We used 4,790 grouse observations (broods and adults) collected over 39,303 stands for 15 years of four forest grouse species (capercaillie, black grouse, hazel grouse, and willow grouse) obtained from comprehensive Finnish wildlife triangle census data and forest variables obtained from Airborne Laser Scanning and satellite data originally sampled at 16 m resolution. We fitted Generalized Additive Mixed Models linking grouse presence/absence in the Finnish boreal forest with forest stand structure and composition. We estimated the effects of predictor variables aggregated at three spatial scales reflecting the species use of the landscape: local level at stand scale, home range level at 1 km radius, and regional level at 5 km radius. Multi-grain models considering forest-species relationships at multiple scales were used to evaluate whether there is a specific scale at which forest characteristics best predict local grouse occupancy. We found that that the spatial scale affected the predictive capacity of the grouse occupancy models and the characteristic scale of habitat selection was the same (i.e., stand scale) among species. Different grouse species exhibited varying optimal spatial scales for occupancy prediction. Forest structure was more important than compositional diversity in predicting grouse occupancy irrespective of the scale. A limited number of forest predictors related to availability of multi-layered vegetation and of suitable thickets explained the occupancy patterns for all the grouse species at different scales. In conclusion, modeling grouse occupancy using forest predictors at different spatial scales can inform forest managers about the scale at which the species perceive the landscape. This evidence calls for an integrated multiscale approach to habitat modelling for forest species.

Deep Weighted Fusion Learning (DWFL)-based multi-sensor fusion model for accurate building occupancy detection

With the advancement of artificial intelligence, the dominance of deep learning (DL) models over ordinary machine learning (ML) algorithms has become a reality in recent years due to its capability of handling complex pattern recognition without manual feature pre-definition. With the growing demands for power savings, building energy loss reduction could benefit from DL techniques. For buildings/rooms with the varying number of occupants, heating, ventilation, and air conditioning (HVAC) systems are often found in operations without much necessity. To reduce the building’s energy loss, accurate occupancy detection/prediction (ODP) results could be used to control the proper operations of HVACs. However, ODP is a challenging issue due to multiple reasons, such as improper selection/deployment of sensors, inefficient learning algorithms for pattern recognition, varying room conditions, etc. To overcome the above challenges, we propose a DL-based framework, i.e., Deep Weighted Fusion Learning (DWFL), to detect and predict occupancy counts with optimal multi-sensor fusion structure. DWFL fuses the extracted features from multiple types of sensors with the priority/weight assignment to each sensor. Such weight assignment considers different room conditions and the pros/cons of each type of sensor. To evaluate DWFL model in terms of occupancy prediction accuracy, we have set up an experimental testbed with low-cost cameras, carbon dioxide (CO2), and passive infrared (PIR) sensors. Among the recently proposed occupancy detection models, DeepFusion utilized deep learning model on heterogeneous sensor data and achieved 88% accuracy in occupancy count estimation (Xue et al., 2019). Another deep learning-based model MI-PIR achieved 91% accuracy on raw analog data from PIR sensors (Andrews et al., 2020). Our research outcome is 94%. Therefore, the experiment results show that our DWFL scheme outperforms the state-of-the-art ODP methods by 3%.