Artificial Neural Networks (ANNs) have become an accepted approach for a wide range of challenges. Meanwhile, the advancement of chip manufacturing processes is approaching saturation which calls for new computing solutions. This work presents a novel approach of an FPGA-based accelerator development for fully connected feed-forward neural networks (FFNNs). A specialized tool was developed to facilitate different implementations, which splits FFNN into elementary layers, allocates computational resources and generates high-level C++ description for high-level synthesis (HLS) tools. Various topologies are implemented and benchmarked, and a comparison with related work is provided. The proposed methodology is applied for the implementation of high-throughput virtual sensor.
The combination of machine learning and heterogeneous embedded platforms enables new potential for developing sophisticated control concepts which are applicable to the field of vehicle dynamics and ADAS. This interdisciplinary work provides enabler solutions -ultimately implementing fast predictions using neural networks (NNs) on field programmable gate arrays (FPGAs) and graphical processing units (GPUs)- while applying them to a challenging application: Torque Vectoring on a multi-electric-motor vehicle for enhanced vehicle dynamics. The foundation motivating this work is provided by discussing multiple domains of the technological context as well as the constraints related to the automotive field, which contrast with the attractiveness of exploiting the capabilities of new embedded platforms to apply advanced control algorithms for complex control problems. In this particular case we target enhanced vehicle dynamics on a multi-motor electric vehicle benefiting from the greater degrees of freedom and controllability offered by such powertrains. Considering the constraints of the application and the implications of the selected multivariable optimization challenge, we propose a NN to provide batch predictions for real-time optimization. This leads to the major contribution of this work: efficient NN implementations on two intrinsically parallel embedded platforms, a GPU and a FPGA, following an analysis of theoretical and practical implications of their different operating paradigms, in order to efficiently harness their computing potential while gaining insight into their peculiarities. The achieved results exceed the expectations and additionally provide a representative illustration of the strengths and weaknesses of each kind of platform. Consequently, having shown the applicability of the proposed solutions, this work contributes valuable enablers also for further developments following similar fundamental principles.
Infrared imaging sensors are frequently used in thermal signature detection applications in industrial, automotive, military and many other areas. However, advanced infrared detectors are generally associated with high costs and complexity. Infrared detectors usually necessitate a thermoelectric heater–cooler for temperature stabilization and various computationally complex preprocessing algorithms for fixed pattern noise (FPN) correction. In this paper, we leverage the benefits of uncooled focal plane arrays and describe a complete digital circuit design for Field Programmable Gate Array (FPGA)-based infrared image acquisition and pre-processing. The proposed design comprises temperature compensation, non-uniformity correction, defective pixel correction cores, spatial image transformation and registration with RGB images. When implemented on Xilinx Ultrascale+ FPGA, the system achieves a throughput of 30 frames per second using the Fraunhofer IMS Digital 17 μm QVGA-IRFPA with a microbolometer array size of 320 × 240 pixels and an RGB camera with a 1024 × 720 resolution. The maximum ratio of the standard deviation to the mean of 0.35% was achieved after FPN correction.
It is predicted that due to such technology-driven trends as electrification, connectivity, autonomous driving and diverse mobility, the automotive market will reach a size of 6.7 trillion USD in 2030. Nevertheless, the composition and architectures of the future autonomous vehicles are still unsettled due to the continuous advancement of sensors and computing hardware, availability and interpretation of new data and safety requirements. Given the existing architectures of the autonomous systems, there is a lack of start-to-end depiction of the self-driving car design and components. This scientific article aspires to serve as a starting guide and boilerplate for the evolution of future architectures of such autonomous systems. The article summarizes the previous research in the field, describes the overall hardware and software composition of the EDI Drive-by-Wire car and investigates the components of the autonomous system: data acquisition, perception algorithms, chosen hardware, software component management, vehicle interfacing, approach to failsafe and fail-operational functionalities.
Technology advancement, major investments, and mainstream interest have reignited the prospect of autonomous driving, nevertheless, the achievement of reliable perception is still an active research topic. This article examines a technique for simplifying path planning and control algorithms by reducing spacial dimensions to two, i.e. by acquiring a Bird’s-Eye View (BEV) image. This article aspires to facilitate novel perception fusion approaches by specifying a geometric approach for inverse perspective mapping (IPM), construction of composite BEV of the vehicle, giving suggestions for implementation and proposing a coherent, BEV-based approach for perception fusion. The proposed geometric IPM approach for BEV image acquisition is suited for simulated environments with limited knowledge of the camera’s intrinsic parameters.
Mechanically scanning LiDAR imaging sensors are abundantly used in applications ranging from basic safety assistance to high-level automated driving, offering excellent spatial resolution and full surround-view coverage in most scenarios. However, their complex optomechanical structure introduces limitations, namely limited mounting options and blind zones, especially in elongated vehicles. To mitigate these challenges, we propose a distributed Time-of-Flight (ToF) sensor system with a flexible hardware–software architecture designed for multi-sensor synchronous triggering and fusion. We formalize the sensor triggering, interference mitigation scheme, data aggregation and fusion procedures and highlight challenges in achieving accurate global registration with current state-of-the-art methods. The resulting surround view visual information is then applied to Spiking Neural Network (SNN)-based object detection and probabilistic occupancy grid mapping (OGM) for enhanced environmental awareness. The proposed system is demonstrated on a test vehicle, achieving coverage of blind zones in a range of 0.5–6 m with a scalable and reconfigurable sensor mounting setup. Using seven ToF sensors, we can achieve a 10 Hz synchronized frame rate, with a 360° point cloud registration and fusion latency below 40 ms. We collected real-world driving data to evaluate the system, achieving 65% mean Average Precision (mAP) in object detection with our SNN. Overall, this work presents a replacement or addition to LiDAR in future high-level automation tasks, offering improved coverage and system integration.