Showing papers on "Obstacle published in 2019"
TL;DR: A critical review on the performance of an obstacle near an exit and a systematic approach of optimising architectural adjustments that enhances escape dynamics of pedestrians’ crowd in indoor and outdoor public spaces needs to be conducted in future.
87 citations
04 Feb 2019
TL;DR: The environment is outlined and a set of baseline results produced by current state-of-the-art Deep RL methods as well as human players are provided; these algorithms fail to produce agents capable of performing near human level.
Abstract: The rapid pace of recent research in AI has been driven in part by the presence of fast and challenging simulation environments. These environments often take the form of games; with tasks ranging from simple board games, to competitive video games. We propose a new benchmark - Obstacle Tower: a high fidelity, 3D, 3rd person, procedurally generated environment. An agent playing Obstacle Tower must learn to solve both low-level control and high-level planning problems in tandem while learning from pixels and a sparse reward signal. Unlike other benchmarks such as the Arcade Learning Environment, evaluation of agent performance in Obstacle Tower is based on an agent's ability to perform well on unseen instances of the environment. In this paper we outline the environment and provide a set of baseline results produced by current state-of-the-art Deep RL methods as well as human players. These algorithms fail to produce agents capable of performing near human level.
84 citations
Posted Content•
TL;DR: The Obstacle Tower as discussed by the authors is a high fidelity, 3D, 3rd person, procedurally generated environment where agents learn to solve both low-level control and high-level planning problems in tandem while learning from pixels and a sparse reward signal.
Abstract: The rapid pace of recent research in AI has been driven in part by the presence of fast and challenging simulation environments. These environments often take the form of games; with tasks ranging from simple board games, to competitive video games. We propose a new benchmark - Obstacle Tower: a high fidelity, 3D, 3rd person, procedurally generated environment. An agent playing Obstacle Tower must learn to solve both low-level control and high-level planning problems in tandem while learning from pixels and a sparse reward signal. Unlike other benchmarks such as the Arcade Learning Environment, evaluation of agent performance in Obstacle Tower is based on an agent's ability to perform well on unseen instances of the environment. In this paper we outline the environment and provide a set of baseline results produced by current state-of-the-art Deep RL methods as well as human players. These algorithms fail to produce agents capable of performing near human level.
80 citations
Posted Content•
TL;DR: It is demonstrated that with the proposed framework, the simulator car agents can transfer knowledge to the RC cars in real-time, with 27% increase in the average distance with obstacles and 42% decrease in the collision counts.
Abstract: Reinforcement learning (RL) is widely used in autonomous driving tasks and training RL models typically involves in a multi-step process: pre-training RL models on simulators, uploading the pre-trained model to real-life robots, and fine-tuning the weight parameters on robot vehicles. This sequential process is extremely time-consuming and more importantly, knowledge from the fine-tuned model stays local and can not be re-used or leveraged collaboratively. To tackle this problem, we present an online federated RL transfer process for real-time knowledge extraction where all the participant agents make corresponding actions with the knowledge learned by others, even when they are acting in very different environments. To validate the effectiveness of the proposed approach, we constructed a real-life collision avoidance system with Microsoft Airsim simulator and NVIDIA JetsonTX2 car agents, which cooperatively learn from scratch to avoid collisions in indoor environment with obstacle objects. We demonstrate that with the proposed framework, the simulator car agents can transfer knowledge to the RC cars in real-time, with 27% increase in the average distance with obstacles and 42% decrease in the collision counts.
68 citations
18 Nov 2019
TL;DR: A novel deep learning-based sensor fusion framework, termed as the “RVNet”, for the effective fusion of the monocular camera and long-range radar for obstacle detection, which is better than baseline algorithms in varying environmental conditions.
Abstract: Camera and radar-based obstacle detection are important research topics in environment perception for autonomous driving. Camera-based obstacle detection reports state-of-the-art accuracy, but the performance is limited in challenging environments. In challenging environments, the camera features are noisy, limiting the detection accuracy. In comparison, the radar-based obstacle detection methods using the 77 GHZ long-range radar are not affected by these challenging environments. However, the radar features are sparse with no delineation of the obstacles. The camera and radar features are complementary, and their fusion results in robust obstacle detection in varied environments. Once calibrated, the radar features can be used for localization of the image obstacles, while the camera features can be used for the delineation of the localized obstacles. We propose a novel deep learning-based sensor fusion framework, termed as the “RVNet”, for the effective fusion of the monocular camera and long-range radar for obstacle detection. The RVNet is a single shot object detection network with two input branches and two output branches. The RVNet input branches contain separate branches for the monocular camera and the radar features. The radar features are formulated using a novel feature descriptor, termed as the “sparse radar image”. For the output branches, the proposed network contains separate branches for small obstacles and big obstacles, respectively. The validation of the proposed network with state-of-the-art baseline algorithm is performed on the Nuscenes public dataset. Additionally, a detailed parameter analysis is performed with several variants of the RVNet. The experimental results show that the proposed network is better than baseline algorithms in varying environmental conditions.
60 citations
TL;DR: In this study, an obstacle avoidance controller based on nonlinear model predictive control is designed in autonomous vehicle navigation and can ensure real-time trajectory tracking and collision avoidance.
Abstract: In this study, an obstacle avoidance controller based on nonlinear model predictive control is designed in autonomous vehicle navigation. The reference trajectory is predefined using a sigmoid function in accordance with road conditions. When obstacles suddenly appear on a predefined trajectory, the reference trajectory should be adjusted dynamically. For dynamic obstacles, a moving trend function is constructed to predict the obstacle position variances in the predictive horizon. Furthermore, a risk index is constructed and introduced into the cost function to realize collision avoidance by combining the relative position relationship between vehicle and obstacles in the predictive horizon. Meanwhile, lateral acceleration constraint is also considered to ensure vehicle stability. Finally, trajectory dynamic planning and tracking are integrated into a single-level model predictive controller. Simulation tests reveal that the designed controller can ensure real-time trajectory tracking and collision avoidance.
59 citations
20 Mar 2019
TL;DR: A new scheme is presented that simultaneously modifies the desired path and speed profile for a vehicle in response to the appearance of an obstacle, significant tracking error, or other environmental change by formulating the trajectory optimization problem as a quadratically constrained quadratic program.
Abstract: In emergency situations, autonomous vehicles will be forced to operate at their friction limits in order to avoid collisions. In these scenarios, coordinating the planning of the vehicle's path and speed gives the vehicle the best chance of avoiding an obstacle. Fast reaction time is also important in an emergency, but approaches to the trajectory planning problem based on nonlinear optimization are computationally expensive. This paper presents a new scheme that simultaneously modifies the desired path and speed profile for a vehicle in response to the appearance of an obstacle, significant tracking error, or other environmental change. By formulating the trajectory optimization problem as a quadratically constrained quadratic program, solution times of less than 20 ms are possible even with a 10-s planning horizon. A simplified point mass model is used to describe the vehicle's motion, but the incorporation of longitudinal weight transfer and road topography means that the vehicle's acceleration limits are modeled more accurately than in comparable approaches. Experimental data from an autonomous vehicle in two scenarios demonstrate how the trajectory planner enables the vehicle to avoid an obstacle even when the obstacle appears suddenly and the vehicle is already operating near the friction limits.
54 citations
20 May 2019
TL;DR: An algorithm is presented that produces a plan for relocating obstacles in order to grasp a target in clutter by a robotic manipulator without collisions and improves up to 31% of the execution time compared to other competitors.
Abstract: We present an algorithm that produces a plan for relocating obstacles in order to grasp a target in clutter by a robotic manipulator without collisions. We consider configurations where objects are densely populated in a constrained and confined space. Thus, there exists no collision-free path for the manipulator without relocating obstacles. Since the problem of planning for object rearrangement has shown to be NP-hard, it is difficult to perform manipulation tasks efficiently which could frequently happen in service domains (e.g., taking out a target from a shelf or a fridge).Our proposed planner employs a collision avoidance scheme which has been widely used in mobile robot navigation. The planner determines an obstacle to be removed quickly in real time. It also can deal with dynamic changes in the configuration (e.g., changes in object poses). Our method is shown to be complete and runs in polynomial time. Experimental results in a realistic simulated environment show that our method improves up to 31% of the execution time compared to other competitors.
50 citations
TL;DR: Examination of the coordination state between urbanization and WRS and its obstacle factors in Beijing city, utilizing the improved coupling coordination degree (ICCD) model, obstacle degree model, and indicator data from 2008 to 2017 indicated an overall upward tendency.
Abstract: Water resource security is an important condition for socio-economic development. Recently, the process of urbanization brings increasing pressures on water resources. Thus, a good understanding of harmonious development of urbanization and water resource security (WRS) systems is necessary. This paper examined the coordination state between urbanization and WRS and its obstacle factors in Beijing city, utilizing the improved coupling coordination degree (ICCD) model, obstacle degree model, and indicator data from 2008 to 2017. Results indicated that: (1) The coupling coordination degree between WRS and urbanization displayed an overall upward tendency during the 2008-2017 period; the coupling coordination state has changed from an imbalanced state into a good coordination state, experiencing from a high-speed development stage (2008-2010), through a steady growth stage (2010-2014), towards a low-speed growth (2014-2017). (2) In urbanization system, both the social and spatial urbanizations have the greatest obstruction to the development of urbanization-WRS system. The subsystems of pressure and state are the domain obstacle subsystems in WRS system. These results can provide important support for urban planning and water resource protection in the future, and hold great significance for urban sustainable development.
46 citations
TL;DR: In this paper, the shape and dimensions of gas flow field channels on the bipolar plates of fuel cells are simulated by selecting the obstacle geometry in the channel path, and after choosing the best range (height), the best obstacle width is analyzed to have better performance.
39 citations
11 Feb 2019
TL;DR: A critical analysis of some of the most promising approaches to geometric collision avoidance in multi-agent systems, namely, the velocity obstacle, reciprocal velocity obstacle (RVO), hybrid-reciprocal velocity obstacles (HRVO), and optimal reciprocal collision avoidance (ORCA) approaches, showing the ORCA method to yield the most scalable computation times and collision likelihood in the presented cases.
Abstract: This paper presents a critical analysis of some of the most promising approaches to geometric collision avoidance in multi-agent systems namely; the velocity obstacle (VO), reciprocal velocity obstacle (RVO), hybrid-reciprocal velocity obstacle (HRVO) and optimal reciprocal collision avoidance (ORCA) approaches. Each approach is evaluated with respect to increasing agent populations and variable sensing assumptions. An intensive 1000 cycle Monte Carlo analysis is used to assess the performance of the selected algorithms in the presented conditions. The optimal reciprocal collision avoidance (ORCA) method is shown to yield the most scalable computation times and collision likelihood in different testing scenarios. The respective features and limitations of each algorithm are discussed and presented through examples.
TL;DR: It is suggested that the non-parallel obstacle layouts (especially the concave layout) with larger longitudinal distance between obstacles are more beneficial to pedestrian movement, when there are obstacles in corridors.
Abstract: Previous research on the effect of obstacles on crowd dynamics primarily focused on relationships between the scale or position of obstacles and pedestrian movement behavior. However, the influence of different obstacle layouts on pedestrian flow in corridors has not been well investigated. Here, we conducted an experimental study on the effect of three obstacle layouts (i.e., parallel, convex and concave layouts) on pedestrian flow in corridors, creating seven scenarios for unidirectional flow and four scenarios for bidirectional flow at low and high crowd densities. The image processing method is employed to collect pedestrian trajectories. Accordingly, typical phenomena, speed, density and flow rate are obtained. It was found that in comparison with the parallel layout, average speed at high crowd density increases approximately 19% in non-parallel layouts when the distance between obstacles is 1.6 m, while the decrease rate of average passing time reaches 17%. The relationship between group speed and time is U-shaped, and group speed is reduced sharply in the parallel layout. Congestion is the severest in the parallel layout in bidirectional flow, i.e. the largest density in the parallel layout is 16% higher than that in non-parallel layouts when the longitudinal distance between obstacles is 1.6 m. Congestion is alleviated with the increasing longitudinal distance between obstacles in the non-parallel layouts. Flow rate in the non-parallel layouts (especially in the concave layout) is greater than that in the parallel layout. These results suggest that the non-parallel obstacle layouts (especially the concave layout) with larger longitudinal distance between obstacles are more beneficial to pedestrian movement, when there are obstacles in corridors. It is hoped that this study will be helpful in crowd management and pedestrian facility design.
TL;DR: The proposed eight-neighbor cells clustering algorithm is used to cluster the obstacle, and static obstacle detection of multi-frame fusion is worked out by combining real-time kinematic global positioning system data and inertial navigation system data of autonomous vehicle.
Abstract: The movement state of obstacle including position, velocity, and yaw angle in the real traffic scenarios has a great impact on the path planning and decision-making of autonomous vehicle. Aiming at...
TL;DR: In this article, the authors proposed a system of nonlinear integral equations based iterative scheme to reconstruct both the location and the shape of the obstacle from the modulus of the far-field data for a single incident plane wave.
Abstract: In this paper, we consider the inverse problem of determining the location and the shape of a sound-soft obstacle from the modulus of the far-field data for a single incident plane wave. By adding a reference ball artificially to the inverse scattering system, we propose a system of nonlinear integral equations based iterative scheme to reconstruct both the location and the shape of the obstacle. The reference ball technique causes few extra computational costs, but breaks the translation invariance and brings information about the location of the obstacle. Several validating numerical examples are provided to illustrate the effectiveness and robustness of the proposed inversion algorithm.
TL;DR: In this article, an empirical kinematic model of the interaction between a soft growing robot and a planar environment is developed to plan paths for the robot to a destination, where the planner exploits obstacle contact when beneficial for navigation.
Abstract: Navigation and motion control of a robot to a destination are tasks that have historically been performed with the assumption that contact with the environment is harmful. This makes sense for rigid-bodied robots where obstacle collisions are fundamentally dangerous. However, because many soft robots have bodies that are low-inertia and compliant, obstacle contact is inherently safe. As a result, constraining paths of the robot to not interact with the environment is not necessary and may be limiting. In this paper, we mathematically formalize interactions of a soft growing robot with a planar environment in an empirical kinematic model. Using this interaction model, we develop a method to plan paths for the robot to a destination. Rather than avoiding contact with the environment, the planner exploits obstacle contact when beneficial for navigation. We find that a planner that takes into account and capitalizes on environmental contact produces paths that are more robust to uncertainty than a planner that avoids all obstacle contact.
TL;DR: It is shown that under the proposed control method, all the robots can always reach into the objective region, maintain their formation, and guarantee collision and obstacle avoidance.
TL;DR: A novel sensor fusion-based system for obstacle detection and identification that uses the millimeter-wave radar to detect the position and velocity of the obstacle and the bounding box regression algorithm in deep learning to precisely locate and identify the obstacles.
Abstract: The purpose of this paper is the presentation and research of a novel sensor fusion-based system for obstacle detection and identification, which uses the millimeter-wave radar to detect the position and velocity of the obstacle. Afterwards, the image processing module uses the bounding box regression algorithm in deep learning to precisely locate and identify the obstacles.,Unlike the traditional algorithms that use radar and vision to detect obstacles separately, the purposed method of this paper uses radar to determine the approximate location of obstacles and then uses bounding box regression to achieve accurate positioning and recognition. First, the information of the obstacles can be acquired by the millimeter-wave radar, and the effective target is extracted by filtering the data. Then, use coordinate system conversion and camera parameter calibration to project the effective target to the image plane, and generate the region of interest (ROI). Finally, based on image processing and machine learning techniques, the vehicle targets in the ROI are detected and tracked.,The millimeter wave is used to determine the presence of an obstacle, and the deep learning algorithm of the image is combined to determine the shape and the class of the obstacle. The experimental results indicate that the detection rate of this method is up to 91.6 per cent, which can better implement the perception of the environment in front of the vehicle.,The originality is based on the combination of millimeter-wave sensors and deep learning. Using the bounding box regression algorithm in RCNN, the ROI detected by radar is analyzed to realize real-time obstacle detection and recognition. This method does not require processing the entire image, greatly reducing the amount of data processing and improving the efficiency of the algorithm.
18 Jul 2019
TL;DR: An autonomous mobile robot has been developed equipped with Light Detection and Ranging (LiDAR) sensor to avoid obstacle and can navigate inside a room without any impact on the wall or the obstacle.
Abstract: In conditions that are dangerous for humans and their environment, the use of robots can be a solution to overcome these problems. Various sensors are used to determine the obstacle free path and the exact position of the robot. However, conventional sensors have limitations in terms of detection distance, spatial resolution, and processing complexity. In this study, an autonomous mobile robot has been developed equipped with Light Detection and Ranging (LiDAR) sensor to avoid obstacle. Braitenberg vehicle strategy is used to navigate the movements of the robot. Sensor data collection and control algorithm are implemented on a single computer board of Raspberry Pi 3. The experimental results show that this sensor can measure distance consistently which is not affected by the object's color and ambient light intensity. The mobile robot can avoid colored objects of different sizes. This autonomous mobile robot can also navigate inside a room without any impact on the wall or the obstacle.
TL;DR: The control scheme proposed in the brief is primarily designed for a communication-free environment where only local state measurements are available and has control input constraints explicitly taken into account.
Abstract: This brief addresses the leader–follower (L-F) tracking control problem for multiple nonholonomic mobile robots in unknown obstacle environments. Unlike most of the existing approaches investigating similar problems, a series of practical issues is considered and tackled in the proposed scheme. For leader tracking, a class of bounded barrier functions are employed to formulate distance and bearing angle constraints introduced by sensor limitations and L-F collision avoidance requirement. To ensure robot safety in unknown environments, a multiregion obstacle avoidance algorithm is proposed which prioritizes different control objectives in different regions. This brief also studies the leader-loss situation, which may be caused by illumination variation, motion blurring, or visual occlusion by obstacles. To deal with this case, a fault-tolerant strategy is designed to drive $F$ to the place where $L$ was lost immediately. The control scheme proposed in the brief is primarily designed for a communication-free environment where only local state measurements are available. Furthermore, it has control input constraints explicitly taken into account. Real robot experiment has been performed to validate the proposed method.
TL;DR: It was found that autonomous control of ground movement and flight was possible for the hybrid aerial / terrestrial robot system, as was autonomous obstacle avoidance by flight when an obstacle appeared during ground movement.
Abstract: To date, many studies related to robots have been performed around the world. Many of these studies have assumed operation at locations where entry is difficult, such as disaster sites, and have focused on various terrestrial robots, such as snake-like, humanoid, spider-type, and wheeled units. Another area of active research in recent years has been aerial robots with small helicopters for operation indoors and outdoors. However, less research has been performed on robots that operate both on the ground and in the air. Accordingly, in this paper, we propose a hybrid aerial / terrestrial robot system. The proposed robot system was developed by equipping a quadcopter with a mechanism for ground movement. It does not use power dedicated to ground movement, and instead uses the flight mechanism of the quadcopter to achieve ground movement as well. Furthermore, we addressed the issue of obstacle avoidance as part of studies on autonomous control. Thus, we found that autonomous control of ground movement and flight was possible for the hybrid aerial / terrestrial robot system, as was autonomous obstacle avoidance by flight when an obstacle appeared during ground movement.
Posted Content•
TL;DR: The feasibility of applying proximal policy optimization, a state-of-the-art deep reinforcement learning algorithm for continuous control tasks, on the dual-objective problem of controlling an underactuated autonomous surface vehicle to follow an a priori known path while avoiding collisions with non-moving obstacles along the way is explored.
Abstract: In this article, we explore the feasibility of applying proximal policy optimization, a state-of-the-art deep reinforcement learning algorithm for continuous control tasks, on the dual-objective problem of controlling an underactuated autonomous surface vehicle to follow an a priori known path while avoiding collisions with non-moving obstacles along the way. The artificial intelligent agent, which is equipped with multiple rangefinder sensors for obstacle detection, is trained and evaluated in a challenging, stochastically generated simulation environment based on the OpenAI gym python toolkit. Notably, the agent is provided with real-time insight into its own reward function, allowing it to dynamically adapt its guidance strategy. Depending on its strategy, which ranges from radical path-adherence to radical obstacle avoidance, the trained agent achieves an episodic success rate between 84 and 100%.
TL;DR: The simulation results showed that the proposed flocking control algorithm provides a better tracking effect and consensus for the UAV swarm when avoiding obstacles.
Abstract: In recent years, with the development of the unmanned aerial vehicle (UAV) and battlefield environments, the UAV swarm has attracted significant research attention. To solve problems regarding poor state consensus among swarm individuals due to a small number of individuals easily falling into local minima upon encountering an obstacle, this paper proposes a flocking obstacle avoidance algorithm with local interaction of obstacle information. To make the UAV swarm follow the desired trajectory with better state consensus, we improved the flocking control algorithm of agents according to the characteristics and requirements of the UAV swarm. The obstacle avoidance algorithm for the UAV swarm is based on Olfati-Saber’s multi-agent obstacle avoidance algorithm. The proposed method has individuals in the swarm communicate obstacle information with their neighbors, and we present a simple analysis of this method. The method improves the cooperative obstacle avoidance capability of the flocking control algorithm. The simulation results showed that the proposed flocking control algorithm provides a better tracking effect and consensus for the UAV swarm when avoiding obstacles.
TL;DR: An improved dynamic virtual ship (DVS) guidance principle is proposed by incorporating the safety prediction obstacle avoidance (SPOA) strategy, which is consistent with the International Regulations of Preventing Collisions at Sea (COLREGs) and the marine practice.
13 May 2019
TL;DR: This paper presents their autonomous vehicle implementation which gives a solid solution to obstacle avoidance in self-driving vehicles in an efficient manner, using a Raspberry Pi and a LIDAR module for indoor navigation.
Abstract: Autonomous miniature vehicles are widely used to test many types of algorithms and simulate the driving behavior as in the real world. In the last decade, several approaches have been proposed for obstacle avoidance in self-driving vehicles. In this paper, we present our autonomous vehicle implementation which gives a solid solution to this problem in an efficient manner, using a Raspberry Pi and a LIDAR module for indoor navigation. Our vehicle is capable of navigating in an unknown environment while avoiding obstacles. Unlike other implementations, our vehicle doesn't use Computer Vision (CV) techniques for obstacle detection but only a single LIDAR sensor and thus can safely navigate in low luminosity enviroments. We perfomed various experiments and verified the success of our implementation.
TL;DR: Practical time-varying output formation tracking problems with collision avoidance, obstacle dodging and connectivity maintenance for high-order multi-agent systems are investigated, and the practical time-Varyingoutput formation tracking error is controlled within an arbitrarily small bound.
Abstract: Practical time-varying output formation tracking problems with collision avoidance, obstacle dodging and connectivity maintenance for high-order multi-agent systems are investigated, and the practical time-varying output formation tracking error is controlled within an arbitrarily small bound. The outputs of followers are designed to track the output of the leader with unknown control input while retaining the predefined time-varying formation. Uncertainties are considered in the dynamics of the followers and the leader. Firstly, distributed extended state observers are developed to estimate the uncertainties and the leader’s unknown control input. A strategy of obstacle dodging is given by designing an ideal secure position for the followers which are in the threatened area of the obstacles. By constructing collision avoidance, obstacle dodging and connectivity maintenance artificial potential functions, corresponding negative gradient terms are calculated to achieve the safety guarantee. Secondly, a practical time-varying output formation tracking protocol is proposed by using distributed extended state observers and the negative gradient terms. Additionally, an approach is presented to determine the gain parameters in the protocol. The stability of the closed-loop multi-agent system with the protocol is analyzed by using Lyapunov stability theory. Finally, a simulation experiment is provided to illustrate the effectiveness of the obtained methods.
25 Sep 2019
TL;DR: A method for formation/collision co-awareness with the goal of energy consumption and response time minimisation is proposed and shows that the proposed approach appropriately keeps the formation during the swarm's travel in the presence of different obstacles.
Abstract: Distributed formation control and obstacle avoidance are two important challenges in autonomous navigation of a swarm of drones and can negatively affect each other due to possible competition that arises between them. In such a platform, a multi-priority control strategy is required to be implemented in every node in order to dynamically optimise the tradeoffs between collision avoidance and formation control w.r.t. given system constraints, e.g. on energy and response time, by reordering priorities in run-time and selecting the suitable formation and collision avoidance approach based on the state of the swarm, i.e., the kinematic parameters and geographical distribution of the nodes as well as the location of the observed obstacles. In this paper, we propose a method for formation/collision co-awareness with the goal of energy consumption and response time minimisation. The algorithm consists of two partial nested feedback-based control loops and based on three observations: 1) for formation maintenance the relative location of the neighbour nodes; 2) observation of an obstacle by a local sensor, represented by a boolean value, used for both formation control and collision avoidance; and 3) in critical situations for avoiding collisions, the distance of an obstacle to the node. The obtained comprehensive experimental results show that the proposed approach appropriately keeps the formation during the swarm's travel in the presence of different obstacles.
TL;DR: This article proposes to defer obstacle resolution to system runtime to meet the system’s goals under changing conditions, and presents techniques for monitoring obstacle satisfaction rates; deciding when adaptation should be triggered; and adapting the system on-the-fly to countermeasures that are more effective.
Abstract: Software systems are deployed in environments that keep changing over time. They should therefore adapt to changing conditions to meet their requirements. The satisfaction rate of these requirements depends on the rate at which adverse conditions prevent their satisfaction. Obstacle analysis is a goal-oriented form of risk analysis for requirements engineering (RE), whereby obstacles to system goals are identified, assessed, and resolved through countermeasures. The selection of effective countermeasures relies on environment assumptions and on the assessed likelihood and criticality of the corresponding obstacles. Those various factors estimated at RE time may, however, evolve at system runtime.To meet the system’s goals under changing conditions, this article proposes to defer obstacle resolution to system runtime. Techniques are presented for monitoring obstacle satisfaction rates; deciding when adaptation should be triggered; and adapting the system on-the-fly to countermeasures that are more effective. The approach relies on a model where goals and obstacles are refined and specified in a probabilistic linear temporal logic. The techniques allow for monitoring the satisfaction rate of probabilistic leaf obstacles; determining the severity of obstacle consequences on goal satisfaction rates computed from the monitored obstacle satisfaction rates; and shifting to countermeasures that better meet the required goal satisfaction rates. Our approach is evaluated on fragments of an ambulance dispatching system.
20 May 2019
TL;DR: This paper presents two novel approaches for improving image-based underwater obstacle detection by combining sparse stereo point clouds with monocular semantic image segmentation and direct binary learning of the presence or absence of underwater obstacles.
Abstract: This paper presents two novel approaches for improving image-based underwater obstacle detection by combining sparse stereo point clouds with monocular semantic image segmentation. Generating accurate image-based obstacle maps in cluttered underwater environments, such as coral reefs, are essential for robust robotic path planning and navigation. However, these maps can be challenged by factors including visibility, lighting and dynamic objects (e.g. fish) that may lead to falsely identified free space or dynamic objects which trajectory planners may react to undesirably. We propose combining feature-based stereo matching with learning-based segmentation to produce a more robust obstacle map. This approach considers direct binary learning of the presence or absence of underwater obstacles, as well as a multiclass learning approach to classify their distance (near, mid and far) in the scene. An enhancement to the binary map is also shown by including depth information from sparse stereo matching to produce 3D obstacle maps of the scene. The performance is evaluated using field data collected in cluttered, and at times, visually degraded coral reef environments. The results show improved image-wide obstacle detection, rejection of transient objects (such as fish), and range estimation compared to feature-based sparse and dense stereo point clouds alone.
TL;DR: A novel proposal assessment method based on fusing statistical properties from both the IPM image and the camera image to perform robust outlier elimination and false positive reduction of generic obstacle detection and collision warning in Advanced Driver Assistance Systems.
Patent•
31 May 2019
TL;DR: In this paper, an intelligent vehicle passable area detection method based on multi-source information fusion is proposed, which comprises the steps: S100, collecting obstacle target information around a vehicle detected by a vehicle-mounted sensor, and outputting a static obstacle target library; S200, receiving obstacle targets around the vehicle, carrying out time-space synchronization on the obstacle targets detected by the vehicle and carrying out single-frame target fusion on all detected obstacle targets, and S300, receiving the static obstacle targets and the dynamic obstacle targets in the step S200 and updating the dynamic
Abstract: The invention discloses an intelligent vehicle passable area detection method based on multi-source information fusion, and the method comprises the steps: S100, collecting obstacle target informationaround a vehicle detected by a vehicle-mounted sensor, and outputting a static obstacle target library; S200, receiving obstacle target information around the vehicle, carrying out time-space synchronization on the obstacle target information detected by the vehicle-mounted sensor, carrying out single-frame target fusion on all detected obstacle information around the vehicle, carrying out continuous inter-frame multi-target tracking by utilizing motion prediction and multi-frame target association, and outputting a dynamic obstacle target library; And S300, receiving the static obstacle target library and the dynamic obstacle target library output in the step S200, and updating the dynamic obstacle target library according to the information of the static obstacle target library to formreal-time obstacle target information and generate a passable area. The position, scale, category and motion information of the obstacle around the vehicle and the binarization rasterized map can be accurately obtained in the vehicle driving process, the motion track of multiple targets is tracked, and the intelligent vehicle passable area including the binarization rasterized map and dynamic obstacle information real-time updating is formed.