Special Issue of the 1st CCF Intelligent Vehicles Symposium, CIVS 2023
Editor's Note
Guest Editor
Article List
2024, 13(2):39-51.
DOI: 10.12146/j.issn.2095-3135.20230724002
Abstract:
To meet the requirements of anti-drone recognition system for drone recognition in the complex background within public places, a target recognition method based on Zebrafish template matching vision recognition and eagle eye visual attention was studied in this paper. By establishing a dataset of drone templates with different postures, combining the eagle eye visual search mechanism with scale invariant feature transformation, the attitude template image is matched with the target to obtain a rough target area. Then calculate the similarity of the Hausdorff distance between the template pose and the target pose to obtain the most similar pose. Experimental results showed that, the anti UAV recognition system can realize the recognition of drones in different complex backgrounds. Compared with the significance target recognition method based on spectral residuals, the average running time is improved by 23.5%. The proposed algorithm has a higher structural similarity index than the differential hash algorithm for finding similar template poses.
To meet the requirements of anti-drone recognition system for drone recognition in the complex background within public places, a target recognition method based on Zebrafish template matching vision recognition and eagle eye visual attention was studied in this paper. By establishing a dataset of drone templates with different postures, combining the eagle eye visual search mechanism with scale invariant feature transformation, the attitude template image is matched with the target to obtain a rough target area. Then calculate the similarity of the Hausdorff distance between the template pose and the target pose to obtain the most similar pose. Experimental results showed that, the anti UAV recognition system can realize the recognition of drones in different complex backgrounds. Compared with the significance target recognition method based on spectral residuals, the average running time is improved by 23.5%. The proposed algorithm has a higher structural similarity index than the differential hash algorithm for finding similar template poses.
2024, 13(2):3-14.
DOI: 10.12146/j.issn.2095-3135.20230817001
Abstract:
In the automatic driving systems, logarithm function has been widely used. For example, logarithm function is often used to design loss function in deep learning or convolutional neural network, which serves as the basis for the automatic driving perception system. Therefore, studying the history of invention of logarithm is of great significance to master the concept and application. This paper studies the definition of Napier’s logarithm and his three tables, analyzes two kinds of proof methods of predecessors, and puts forward new proof methods based on the exponential function. Meanwhile, this paper also analyzes Napier’s calculation method. Compared with other alternative methods, the optimization results of Napier’s interval approximation are given. The calculation by MPRF library shows that Napier’s method is more convergent to the true value.
In the automatic driving systems, logarithm function has been widely used. For example, logarithm function is often used to design loss function in deep learning or convolutional neural network, which serves as the basis for the automatic driving perception system. Therefore, studying the history of invention of logarithm is of great significance to master the concept and application. This paper studies the definition of Napier’s logarithm and his three tables, analyzes two kinds of proof methods of predecessors, and puts forward new proof methods based on the exponential function. Meanwhile, this paper also analyzes Napier’s calculation method. Compared with other alternative methods, the optimization results of Napier’s interval approximation are given. The calculation by MPRF library shows that Napier’s method is more convergent to the true value.
2024, 13(2):15-28.
DOI: 10.12146/j.issn.2095-3135.20230726001
Abstract:
In the field of autonomous driving safety research and application, the limitations of limited testing mileage and exposure to only a single hazardous scenario hinder the improvement of autonomous driving safety performance. To address these issues, testing with adversarial scenarios is considered crucial. However, existing studies utilize generic optimization algorithms as frameworks, resulting in a wastage of computational resources in exploring the parameter space, thereby leading to low efficiency. Moreover, under the constraint of computational cost, these algorithms may not be able to test a sufficient number of diverse failure samples, especially in complex environments. Adversarial scenario testing in complex environments faces three major challenges: information scarcity, sparse distribution of adversarial samples in a vast parameter space, and the difficulty in balancing exploration and exploitation during the search process. To tackle these challenges, this paper proposes an efficient framework for adversarial scenario testing. This framework employs a surrogate model to gather more information about the parameter space, selects small samples to overcome the sparse event constraints in the vast space, and focuses on the unknown regions and adversarial samples for targeted search and update, thereby achieving a balance between exploration and exploitation. Experimental results demonstrate that the proposed method in this paper exhibits a search efficiency four times higher than random sampling and more than double the efficiency compared to general genetic algorithms. Additionally, with a limited number of simulation test runs, it generates a greater number of adversarial test cases that are likely to cause the tested autonomous driving system to fail. Notably, the proposed method can identify many outlier adversarial samples, unveiling failure modes that existing algorithms fail to recognize. Furthermore, the proposed method can swiftly and comprehensively identify the vulnerable scenarios of the tested algorithm, providing support for the testing, validation, and iterative upgrade of autonomous driving algorithms.
In the field of autonomous driving safety research and application, the limitations of limited testing mileage and exposure to only a single hazardous scenario hinder the improvement of autonomous driving safety performance. To address these issues, testing with adversarial scenarios is considered crucial. However, existing studies utilize generic optimization algorithms as frameworks, resulting in a wastage of computational resources in exploring the parameter space, thereby leading to low efficiency. Moreover, under the constraint of computational cost, these algorithms may not be able to test a sufficient number of diverse failure samples, especially in complex environments. Adversarial scenario testing in complex environments faces three major challenges: information scarcity, sparse distribution of adversarial samples in a vast parameter space, and the difficulty in balancing exploration and exploitation during the search process. To tackle these challenges, this paper proposes an efficient framework for adversarial scenario testing. This framework employs a surrogate model to gather more information about the parameter space, selects small samples to overcome the sparse event constraints in the vast space, and focuses on the unknown regions and adversarial samples for targeted search and update, thereby achieving a balance between exploration and exploitation. Experimental results demonstrate that the proposed method in this paper exhibits a search efficiency four times higher than random sampling and more than double the efficiency compared to general genetic algorithms. Additionally, with a limited number of simulation test runs, it generates a greater number of adversarial test cases that are likely to cause the tested autonomous driving system to fail. Notably, the proposed method can identify many outlier adversarial samples, unveiling failure modes that existing algorithms fail to recognize. Furthermore, the proposed method can swiftly and comprehensively identify the vulnerable scenarios of the tested algorithm, providing support for the testing, validation, and iterative upgrade of autonomous driving algorithms.
2024, 13(2):29-38.
DOI: 10.12146/j.issn.2095-3135.20230724001
Abstract:
Chaos LiDAR has attracted significant attention due to its high resolution, inherent antiinterference capability, and stealth characteristics. However, the performance of chaos LiDAR in longrange target detection and imaging is quite limited by the power of chaotic light sources, sensitivity of linear detectors, and hardware bandwidth. To overcome the bottleneck of chaos LiDAR, this paper proposes the concept of digital chaos LiDAR and conducts theoretical analysis and simulation verification. Through Monte Carlo simulation, this paper studied the detection probability, false probability, and detection range of continuous-wave chaos LiDAR, pulsed chaos LiDAR, and digital chaos LiDAR. The simulation results show that, within the confidence interval where the detection probability is greater than 95% and the false alarm probability is less than 5%, the detection range of digital chaos LiDAR is approximately 35 times and 8 times higher than that of continuous-wave chaos LiDAR and pulsed chaos LiDAR, respectively. With the advantages of ultra-high sensitivity of single-photon detectors and digital output, digital chaos LiDAR is expected to be widely used in the field of long-range target detection and imaging.
Chaos LiDAR has attracted significant attention due to its high resolution, inherent antiinterference capability, and stealth characteristics. However, the performance of chaos LiDAR in longrange target detection and imaging is quite limited by the power of chaotic light sources, sensitivity of linear detectors, and hardware bandwidth. To overcome the bottleneck of chaos LiDAR, this paper proposes the concept of digital chaos LiDAR and conducts theoretical analysis and simulation verification. Through Monte Carlo simulation, this paper studied the detection probability, false probability, and detection range of continuous-wave chaos LiDAR, pulsed chaos LiDAR, and digital chaos LiDAR. The simulation results show that, within the confidence interval where the detection probability is greater than 95% and the false alarm probability is less than 5%, the detection range of digital chaos LiDAR is approximately 35 times and 8 times higher than that of continuous-wave chaos LiDAR and pulsed chaos LiDAR, respectively. With the advantages of ultra-high sensitivity of single-photon detectors and digital output, digital chaos LiDAR is expected to be widely used in the field of long-range target detection and imaging.
2024, 13(2):52-63.
DOI: 10.12146/j.issn.2095-3135.20230717001
Abstract:
To realize efficient path exploration of unmanned platforms in unknown environments, a path planning algorithm based on a hierarchical architecture of “perception planning control” is studied in this work. Real-time construction of two-dimensional grid maps of unknown environments using the Cartographer mapping algorithm at the perception layer. At the planning level, the optimal exploration target point is selected by Canny edge detection, density-based clustering algorithm, and performance function evaluation. Specifically, the concept of continuity of exploration direction is introduced into the efficiency function of planning, overcoming the drawbacks of traditional path planning that repeatedly explores known environments. At the control layer, the shortest path from the current pose to the target point is planned using probability roadmap algorithm, and collision-free tracking of the path is achieved through pure tracking algorithm and vector histogram algorithm. The effectiveness of the algorithm was verified through simulation in three typical environments, and the results showed that the proposed algorithm can achieve higher exploration efficiency and completeness in different environments.
To realize efficient path exploration of unmanned platforms in unknown environments, a path planning algorithm based on a hierarchical architecture of “perception planning control” is studied in this work. Real-time construction of two-dimensional grid maps of unknown environments using the Cartographer mapping algorithm at the perception layer. At the planning level, the optimal exploration target point is selected by Canny edge detection, density-based clustering algorithm, and performance function evaluation. Specifically, the concept of continuity of exploration direction is introduced into the efficiency function of planning, overcoming the drawbacks of traditional path planning that repeatedly explores known environments. At the control layer, the shortest path from the current pose to the target point is planned using probability roadmap algorithm, and collision-free tracking of the path is achieved through pure tracking algorithm and vector histogram algorithm. The effectiveness of the algorithm was verified through simulation in three typical environments, and the results showed that the proposed algorithm can achieve higher exploration efficiency and completeness in different environments.
2024, 13(2):64-73.
DOI: 10.12146/j.issn.2095-3135.20230725001
Abstract:
The paper explores the formation control problem of multiple unmanned intelligent vehicles moving in complex environments with the leader-follower method, and designs a formation controller and a formation control scheme by adopting a closed-loop control law, which is advantageous in that it realizes precise control mainly by considering the distance and angle between intelligent vehicles, while referring to the information between the leader and the neighboring followers. Based on the simulated test environment built, the improved control method is tested against the traditional formation method. The experimental results show that the method proposed in this paper has better motion control effects in complex environments.
The paper explores the formation control problem of multiple unmanned intelligent vehicles moving in complex environments with the leader-follower method, and designs a formation controller and a formation control scheme by adopting a closed-loop control law, which is advantageous in that it realizes precise control mainly by considering the distance and angle between intelligent vehicles, while referring to the information between the leader and the neighboring followers. Based on the simulated test environment built, the improved control method is tested against the traditional formation method. The experimental results show that the method proposed in this paper has better motion control effects in complex environments.
