In this paper, one solution for an end-to-end deep neural network for autonomous driving is presented. The main objective of our work was to achieve autonomous driving with a light deep neural network suitable for deployment on embedded automotive platforms. There are several end-to-end deep neural networks used for autonomous driving, where the input to the machine learning algorithm are camera images and the output is the steering angle prediction, but those convolutional neural networks are significantly more complex than the network architecture we are proposing. The network architecture, computational complexity, and performance evaluation during autonomous driving using our network are compared with two other convolutional neural networks that we re-implemented with the aim to have an objective evaluation of the proposed network. The trained model of the proposed network is four times smaller than the PilotNet model and about 250 times smaller than AlexNet model. While complexity and size of the novel network are reduced in comparison to other models, which leads to lower latency and higher frame rate during inference, our network maintained the performance, achieving successful autonomous driving with similar efficiency compared to autonomous driving using two other models. Moreover, the proposed deep neural network downsized the needs for real-time inference hardware in terms of computational power, cost, and size.

A new method for estimation of angles of leg segments and joints, which uses accelerometer arrays attached to body segments, is described. An array consists of two accelerometers mounted on a rigid rod. The absolute angle of each body segment was determined by band pass filtering of the differences between signals from parallel axes from two accelerometers mounted on the same rod. Joint angles were evaluated by subtracting absolute angles of the neighboring segments. This method eliminates the need for double integration as well as the drift typical for double integration. The efficiency of the algorithm is illustrated by experimental results involving healthy subjects who walked on a treadmill at various speeds, ranging between 0.15 m/s and 2.0 m/s. The validation was performed by comparing the estimated joint angles with the joint angles measured with flexible goniometers. The discrepancies were assessed by the differences between the two sets of data (obtained to be below 6 degrees) and by the Pearson correlation coefficient (greater than 0.97 for the knee angle and greater than 0.85 for the ankle angle).

We propose a novel simple method for quantitative and qualitative finger-tapping assessment based on miniature inertial sensors (3D gyroscopes) placed on the thumb and index-finger. We propose a simplified description of the finger tapping by using a single angle, describing rotation around a dominant axis. The method was verified on twelve subjects, who performed various tapping tasks, mimicking impaired patterns. The obtained tapping angles were compared with results of a motion capture camera system, demonstrating excellent accuracy. The root-mean-square (RMS) error between the two sets of data is, on average, below 4°, and the intraclass correlation coefficient is, on average, greater than 0.972. Data obtained by the proposed method may be used together with scores from clinical tests to enable a better diagnostic. Along with hardware simplicity, this makes the proposed method a promising candidate for use in clinical practice. Furthermore, our definition of the tapping angle can be applied to all tapping assessment systems.

The control of movement in humans is hierarchical and distributed and uses feedback. An assistive system could be best integrated into the therapy of a human with a central nervous system lesion if the system is controlled in a similar manner. Here, we present a novel wireless architecture and routing protocol for a distributed functional electrical stimulation system that enables control of movement.

This paper deals with central powering of very low power sensor networks deployed over wide areas. Although wireless communications are nowadays dominant in the majority of implementations, wired differential communications sometimes provide the only ultimately reliable solution. In a large-scale network it is of interest to implement both communication and powering with a small number of thin conductors. The consequential problem of the serial resistance of the long powering lines appears as the limiting factor. Sensor nodes powered with linear supply support very low power consumption, but the current flow through ground loops creates common mode voltage on communication lines. An introduction of the galvanic isolation resolves the common mode issue, but the isolation usually consumes more power than a sensor node itself, which in total gives even worse results than linear regulation. Moreover, a galvanic isolation demands an additional reference conductor for signal return path. In this paper, a novel approach to the realization of a floating linear power supply is proposed, incorporating advantages of both linear and isolated solutions. The new power supply scheme retains linear regulation characteristics regarding low power consumption, at the same time eliminating common mode voltage issue, like an isolated power supply.

This paper presents the analysis of the J-Net deep neural network used for autonomous vehicle lane keeping, and its verification in simulated and real conditions. The transition from the simulated world on the personal computer to the real world in laboratory conditions is tackled. In the presented solution, autonomous lane keeping is achieved by analyzing information from visual sensors using a deep neural network. J-Net was developed with an aim to be implemented on an autonomous vehicle platform with limited hardware performance in terms of computing power and memory capacity. Verification of autonomous driving using J-Net was achieved in simulated conditions, using an open-source simulator for autonomous driving, and in real-world conditions. For the verification in real-world conditions, an autonomous driving system was designed and implemented in the Laboratory of Electronics at the School of Electrical Engineering, University of Belgrade.