Human drivers naturally reason about interactions between road users to understand and safely navigate through traffic. Thus, developing autonomous vehicles necessitates the ability to mimic such knowledge and model interactions between road users to understand and navigate unpredictable, dynamic environments. However, since real-world scenarios often differ from training datasets, effectively modeling the behavior of various road users in an environment remains a significant research challenge. This reality necessitates models that generalize to a broad range of domains and explicitly model interactions between road users and the environment to improve scenario understanding. Graph learning methods address this problem by modeling interactions using graph representations of scenarios. However, existing methods cannot effectively transfer knowledge gained from the training domain to real-world scenarios. This constraint is caused by the domain-specific rules used for graph extraction that can vary in effectiveness across domains, limiting generalization ability. To address these limitations, we propose RoadScene2Graph (RS2G): a data-driven graph extraction and modeling approach that learns to extract the best graph representation of a road scene for solving autonomous scene understanding tasks. We show that RS2G enables better performance at subjective risk assessment than rule-based graph extraction methods and deep-learning-based models. RS2G also improves generalization and Sim2Real transfer learning, which denotes the ability to transfer knowledge gained from simulation datasets to unseen real-world scenarios. We also present ablation studies showing how RS2G produces a more useful graph representation for downstream classifiers. Finally, we show how RS2G can identify the relative importance of rule-based graph edges and enables intelligent graph sparsity tuning.
Wearable medical technology has become increasingly popular in recent years. One function of wearable health devices is stress detection, which relies on sensor inputs to determine the mental state of patients. This continuous, real-time monitoring can provide healthcare professionals with vital physiological data and enhance the quality of patient care. Current methods of stress detection lack: (i) robustness -- wearable health sensors contain high levels of measurement noise that degrades performance, and (ii) adaptation -- static architectures fail to adapt to changing contexts in sensing conditions. We propose to address these deficiencies with SELF-CARE, a generalized selective sensor fusion method of stress detection that employs novel techniques of context identification and ensemble machine learning. SELF-CARE uses a learning-based classifier to process sensor features and model the environmental variations in sensing conditions known as the noise context. SELF-CARE uses noise context to selectively fuse different sensor combinations across an ensemble of models to perform robust stress classification. Our findings suggest that for wrist-worn devices, sensors that measure motion are most suitable to understand noise context, while for chest-worn devices, the most suitable sensors are those that detect muscle contraction. SELF-CARE demonstrates state-of-the-art performance on the WESAD dataset. Using wrist-based sensors, SELF-CARE achieves 86.34% and 94.12% accuracy for the 3-class and 2-class stress classification problems, respectively. For chest-based wearable sensors, SELF-CARE achieves 86.19% (3-class) and 93.68% (2-class) classification accuracy. This work demonstrates the benefits of utilizing selective, context-aware sensor fusion in mobile health sensing that can be applied broadly to Internet of Things applications.
Thanks to the rapid growth in wearable technologies and recent advancement in machine learning and signal processing, monitoring complex human contexts becomes feasible, paving the way to develop human-in-the-loop IoT systems that naturally evolve to adapt to the human and environment state autonomously. Nevertheless, a central challenge in designing many of these IoT systems arises from the requirement to infer the human mental state, such as intention, stress, cognition load, or learning ability. While different human contexts can be inferred from the fusion of different sensor modalities that can correlate to a particular mental state, the human brain provides a richer sensor modality that gives us more insights into the required human context. This paper proposes ERUDITE, a human-in-the-loop IoT system for the learning environment that exploits recent wearable neurotechnology to decode brain signals. Through insights from concept learning theory, ERUDITE can infer the human state of learning and understand when human learning increases or declines. By quantifying human learning as an input sensory signal, ERUDITE can provide adequate personalized feedback to humans in a learning environment to enhance their learning experience. ERUDITE is evaluated across $15$ participants and showed that by using the brain signals as a sensor modality to infer the human learning state and providing personalized adaptation to the learning environment, the participants' learning performance increased on average by $26\%$. Furthermore, we showed that ERUDITE can be deployed on an edge-based prototype to evaluate its practicality and scalability.
Heterogeneous MPSoCs comprise diverse processing units of varying compute capabilities. To date, the mapping strategies of neural networks (NNs) onto such systems are yet to exploit the full potential of processing parallelism, made possible through both the intrinsic NNs' structure and underlying hardware composition. In this paper, we propose a novel framework to effectively map NNs onto heterogeneous MPSoCs in a manner that enables them to leverage the underlying processing concurrency. Specifically, our approach identifies an optimal partitioning scheme of the NN along its `width' dimension, which facilitates deployment of concurrent NN blocks onto different hardware computing units. Additionally, our approach contributes a novel scheme to deploy partitioned NNs onto the MPSoC as dynamic multi-exit networks for additional performance gains. Our experiments on a standard MPSoC platform have yielded dynamic mapping configurations that are 2.1x more energy-efficient than the GPU-only mapping while incurring 1.7x less latency than DLA-only mapping.
Runtime energy management has become quintessential for multi-sensor autonomous systems at the edge for achieving high performance given the platform constraints. Typical for such systems, however, is to have their controllers designed with formal guarantees on safety that precede in priority such optimizations, which in turn limits their application in real settings. In this paper, we propose a novel energy optimization framework that is aware of the autonomous system's safety state, and leverages it to regulate the application of energy optimization methods so that the system's formal safety properties are preserved. In particular, through the formal characterization of a system's safety state as a dynamic processing deadline, the computing workloads of the underlying models can be adapted accordingly. For our experiments, we model two popular runtime energy optimization methods, offloading and gating, and simulate an autonomous driving system (ADS) use-case in the CARLA simulation environment with performance characterizations obtained from the standard Nvidia Drive PX2 ADS platform. Our results demonstrate that through a formal awareness of the perceived risks in the test case scenario, energy efficiency gains are still achieved (reaching 89.9%) while maintaining the desired safety properties.
To mitigate the high energy demand of Neural Network (NN) based Autonomous Driving Systems (ADSs), we consider the problem of offloading NN controllers from the ADS to nearby edge-computing infrastructure, but in such a way that formal vehicle safety properties are guaranteed. In particular, we propose the EnergyShield framework, which repurposes a controller ''shield'' as a low-power runtime safety monitor for the ADS vehicle. Specifically, the shield in EnergyShield provides not only safety interventions but also a formal, state-based quantification of the tolerable edge response time before vehicle safety is compromised. Using EnergyShield, an ADS can then save energy by wirelessly offloading NN computations to edge computers, while still maintaining a formal guarantee of safety until it receives a response (on-vehicle hardware provides a just-in-time fail safe). To validate the benefits of EnergyShield, we implemented and tested it in the Carla simulation environment. Our results show that EnergyShield maintains safe vehicle operation while providing significant energy savings compared to on-vehicle NN evaluation: from 24% to 54% less energy across a range of wireless conditions and edge delays.
Dynamic neural networks (DyNNs) have become viable techniques to enable intelligence on resource-constrained edge devices while maintaining computational efficiency. In many cases, the implementation of DyNNs can be sub-optimal due to its underlying backbone architecture being developed at the design stage independent of both: (i) the dynamic computing features, e.g. early exiting, and (ii) the resource efficiency features of the underlying hardware, e.g., dynamic voltage and frequency scaling (DVFS). Addressing this, we present HADAS, a novel Hardware-Aware Dynamic Neural Architecture Search framework that realizes DyNN architectures whose backbone, early exiting features, and DVFS settings have been jointly optimized to maximize performance and resource efficiency. Our experiments using the CIFAR-100 dataset and a diverse set of edge computing platforms have seen HADAS dynamic models achieve up to 57% energy efficiency gains compared to the conventional dynamic ones while maintaining the desired level of accuracy scores. Our code is available at https://github.com/HalimaBouzidi/HADAS
Due to the high performance and safety requirements of self-driving applications, the complexity of modern autonomous driving systems (ADS) has been growing, instigating the need for more sophisticated hardware which could add to the energy footprint of the ADS platform. Addressing this, edge computing is poised to encompass self-driving applications, enabling the compute-intensive autonomy-related tasks to be offloaded for processing at compute-capable edge servers. Nonetheless, the intricate hardware architecture of ADS platforms, in addition to the stringent robustness demands, set forth complications for task offloading which are unique to autonomous driving. Hence, we present $ROMANUS$, a methodology for robust and efficient task offloading for modular ADS platforms with multi-sensor processing pipelines. Our methodology entails two phases: (i) the introduction of efficient offloading points along the execution path of the involved deep learning models, and (ii) the implementation of a runtime solution based on Deep Reinforcement Learning to adapt the operating mode according to variations in the perceived road scene complexity, network connectivity, and server load. Experiments on the object detection use case demonstrated that our approach is 14.99% more energy-efficient than pure local execution while achieving a 77.06% reduction in risky behavior from a robust-agnostic offloading baseline.
The globalization of the Integrated Circuit (IC) supply chain has moved most of the design, fabrication, and testing process from a single trusted entity to various untrusted third-party entities worldwide. The risk of using untrusted third-Party Intellectual Property (3PIP) is the possibility for adversaries to insert malicious modifications known as Hardware Trojans (HTs). These HTs can compromise the integrity, deteriorate the performance, deny the service, and alter the functionality of the design. While numerous HT detection methods have been proposed in the literature, the crucial task of HT localization is overlooked. Moreover, a few existing HT localization methods have several weaknesses: reliance on a golden reference, inability to generalize for all types of HT, lack of scalability, low localization resolution, and manual feature engineering/property definition. To overcome their shortcomings, we propose a novel, golden reference-free HT localization method at the pre-silicon stage by leveraging Graph Convolutional Network (GCN). In this work, we convert the circuit design to its intrinsic data structure, graph and extract the node attributes. Afterward, the graph convolution performs automatic feature extraction for nodes to classify the nodes as Trojan or benign. Our automated approach does not burden the designer with manual code review. It locates the Trojan signals with 99.6% accuracy, 93.1% F1-score, and a false-positive rate below 0.009%.
Providing health monitoring devices with machine intelligence is important for enabling automatic mobile healthcare applications. However, this brings additional challenges due to the resource scarcity of these devices. This work introduces a neural contextual bandits based dynamic sensor selection methodology for high-performance and resource-efficient body-area networks to realize next generation mobile health monitoring devices. The methodology utilizes contextual bandits to select the most informative sensor combinations during runtime and ignore redundant data for decreasing transmission and computing power in a body area network (BAN). The proposed method has been validated using one of the most common health monitoring applications: cardiac activity monitoring. Solutions from our proposed method are compared against those from related works in terms of classification performance and energy while considering the communication energy consumption. Our final solutions could reach $78.8\%$ AU-PRC on the PTB-XL ECG dataset for cardiac abnormality detection while decreasing the overall energy consumption and computational energy by $3.7 \times$ and $4.3 \times$, respectively.