In many service applications, mobile robots need to share their work areas with obstacles. Avoiding collisions is a fundamental requirement for these robots. In this paper a novel collision avoidance ...system is developed for avoiding unpredictable dynamic obstacles, including humans. The collision avoidance algorithm is based on the virtual force field (VFF) concept. The velocities of the obstacles are used in addition to their positions to improve the avoidance performance for dynamic obstacles. Unlike prior algorithms, the proposed VFF is designed to be continuous to diminish both path oscillations and the time cost for reaching the goal. To further reduce the time cost, a new virtual force (termed the detour force) is introduced. The detour force also solves the challenging avoidance problem that occurs when the centers of the robot, human/obstacle and goal are collinear; and the human/obstacle and robot are moving towards each other. In simulations and experiments with a maximum approach velocity of 1.7 m/s, the avoidance system with the new VFF algorithm generates collision-free paths with less oscillation and lower time cost.
Unintentional physical human-robot contact is becoming more common as robots operate in closer proximity to people. This contact may generate a large impact force and cause severe human injuries. ...Therefore, the ability to reduce the human-robot impact force and ensure human safety is a fundamental requirement for human-friendly robots. An easy and effective way to achieve this is using foam to cover the robot surface. We present a method for designing the stiffness and thickness of the foam covering based on a realistic safety threshold and an improved impact force model. Our model incorporates the previously neglected coupling of the human head to the torso and the coupling of the robot arm to its base. The impact model and model-based design procedure are experimentally verified for various foam properties, and robot and human velocities. The impact experiments are performed with an apparatus simulating the human head and, at lower velocity, with a human volunteer. The maximum error between the predicted and experimental peak impact force results is 8%.
This paper presents three algebraic methods for constructing nonbinary quasi-cyclic (QC)-LDPC codes. Three classes of efficiently encodable QC-LDPC codes over nonbinary finite fields are constructed. ...Experimental results show that constructed codes decoded with iterative decoding perform well over the AWGN channel and they achieve significant coding gains over Reed-Solomon (RS) codes of the same lengths and rates decoded with algebraic decoding
Mitosis detection is one of the challenging problems in computational pathology, and mitotic count is an important index of cancer grading for pathologists. However, current counts of mitotic nuclei ...rely on pathologists looking microscopically at the number of mitotic nuclei in hot spots, which is subjective and time-consuming. In this paper, we propose a two-stage cascaded network, named FoCasNet, for mitosis detection. In the first stage, a detection network named M_det is proposed to detect as many mitoses as possible. In the second stage, a classification network M_class is proposed to refine the results of the first stage. In addition, the attention mechanism, normalization method, and hybrid anchor branch classification subnet are introduced to improve the overall detection performance. Our method achieves the current highest F1-score of 0.888 on the public dataset ICPR 2012. We also evaluated our method on the GZMH dataset released by our research team for the first time and reached the highest F1-score of 0.563, which is also better than multiple classic detection networks widely used at present. It confirmed the effectiveness and generalization of our method. The code will be available at: https://github.com/antifen/mitosis-nuclei-detection.
We propose a novel iterative decoding algorithm based on partial combination of the parity check matrix for iterative soft decoding of Reed-Solomon product codes. It uses only the simple syndrome ...equations whose solutions are confined to parts of the columns of the parity check matrix. Compared to other algorithms, the proposed algorithm has lower complexity while offering better performance, which is demonstrated by simulations.
This paper presents three algebraic methods for constructing q-ary LDPC codes. The first method gives a class of dispersed Reed-Solomon codes as LDPC codes. The second method gives a class of q-ary ...quasi-cyclic LDPC codes. The third method gives two classes of q-ary finite geometry LDPC codes. Codes constructed by these methods perform very well with iterative decoding, even for short codes.
This dissertation presents three algebraic methods for constructing nonbinary low-density parity-check (LDPC) codes whose Tanner graphs has girth at least 6. The experimental results show that these ...codes perform well under iterative decoding algorithm. Compared with other nonbinary LDPC codes, codes constructed by algebraic methods have quasi-cyclic or cyclic structure and therefore allow efficient encoding. First presented is a finite field approach for constructing two classes of quasi-cyclic LDPC codes. The parity-check matrices of the codes constructed by the finite field approach usually have full row rank or nearly full row rank. Hence, the encoding complexity is small. In general, this approach is suitable for constructing high-rate codes, whose parity-check matrices have small column weights. Next a finite geometry approach is presented for constructing one class of cyclic LDPC codes, three classes of quasi-cyclic LDPC codes and one class of structured regular LDPC codes. The parity-check matrices of the codes constructed by the finite geometry approach usually have large column weights, hence these codes may show a very low error floor. Iterative decoding of these nonbinary LDPC codes converges very fast. Then a superposition-dispersion method is devised for constructing long quasi-cyclic LDPC codes from short codes with small symbol size. The short codes can be constructed by the two previous approaches. Finally, the efficient encoding of quasi-cyclic LDPC codes using shift registers is presented.
This paper presents methods for efficient encoding of quasi-cyclic LDPC codes. Based on these methods, encoding of quasi-cyclic LDPC codes can be implemented using simple shift-registers with ...complexity linearly proportional to the number of parity-check bits of a code for serial encoding and to the length of a code for parallel encoding. Various encoding circuits are devised and they provide a range of trade-offs between encoding complexity and speed.
In this thesis, solutions to two of the problems encountered in the design and control of human-friendly robots are investigated. The first problem is severe human injuries can occur when an ...accidental human-manipulator impact happens. A theoretical and experimental study on using foam coverings to reduce the severity of a human-manipulator impact and enhance human safety is presented. An improved human-manipulator impact model that incorporates the manipulator dynamics, foam covering dynamics and the coupling between the human head and torso is introduced. A method for approximating the configuration-dependent dynamics of robotics manipulators with the dynamics of a single DOF manipulator is proposed. With this model, the design parameters that significantly influence the human head acceleration are investigated. A model-based foam covering design procedure to properly select parameters of foam coverings in accordance with safety criteria and the foam thickness constraint is then proposed. The impact model and the foam covering design procedure are validated experimentally with two manipulators. The maximum error between the predicted and experimental head acceleration was less than 9%. The maximum error between the;
predicted and experimental foam compressed depth was less than 12%. The second problem is mobile robot navigation in the presence of humans and other motion-unpredictable obstacles. A novel navigation algorithm, based on the virtual force field (VFF) method, is proposed as a solution. It features improved functions for the repulsive and detour virtual forces, and a new stabilizing virtual force. Methods to calculate sizes of the active and critical regions for different obstacles are developed. Stability of the new VFF is proven using a novel piecewise Lyapunov function and Lyapunov's second method. Based on simulations for different obstacle configurations, the new VFF-based algorithm successfully produces collision-free paths while five well known navigation algorithms incurred collisions in one of the configurations. With the new VFF-based navigation algorithm, simulations and experiments are successfully performed with a holonomic robot and a nonholonomic robot for several configurations, including multiple moving obstacles.
Thesis
Doctor of Philosophy (PhD)
The development of a system for automatically locating and tracking a human in the vicinity of a robot is described. The system consists of multiple passive infrared (PIR) sensors, two color cameras, ...a pair of microwave sensors and a pair of PCs for data collection, signal processing and data fusion. The cameras are treated as individual sensors rather than a stereo pair to minimize the affect of occlusion by the robot. The area around the robot is subdivided into an occupancy grid with 0.5m by 0.5m cells. A data fusion algorithm, based on Dempster-Shafer evidence theory, is used to estimate the probability of human occupancy for each cell. This information is used to estimate the human's location. A novel concept termed a "protective cellâ⬠is introduced to further increase the human's safety in the presence of sensor uncertainty. Experimental results are included demonstrating the system's effectiveness.