China
Africa
Explore chapters and articles related to this topic
Application of Artificial Intelligence Algorithms for Robot Development
Published in S. S. Nandhini, M. Karthiga, S. B. Goyal, Computational Intelligence in Robotics and Automation, 2023
R. M. Tharsanee, R. S. Soundariya, A. Saran Kumar, V. Praveen
Path planning is one of the important problems faced by the robots. Robots are designed with several sensors for vision like camera, and sound recognition is aided by ultrasonic sensors. The data extracted from these sensors are not high-level information or not semantic in nature. With sensors, robots can determine an object which lies a few feet away from them, but it is highly difficult for the robot to make decisions on how to react without knowing the type of object. Neural networks can be very useful for the extraction of high-order information from the low-order non-semantic data. In robot navigation, robot is expected to follow a path to the destination by avoiding the objects that are on the way and reach the goal successfully. Using neural networks for path planning involves three major steps. First, the robot should be fed with a complete knowledge of the environment. Second, it should be capable enough to identify the movements required from the source location to the destination location inside the environment. But it is not possible to identify the entire movable and static objects in the environment beforehand. The real challenge occurs in the last step which can be solved by the planning methods that can be implemented in the local environment.
Robotic Path-Planning in Dynamic and Uncertain Environment Using Genetic Algorithm
Published in Jitendra R. Raol, Ajith K. Gopal, Mobile Intelligent Autonomous Systems, 2016
Robotic navigation encompasses (1) motion-planning which includes dynamical modelling and (2) path-planning which restricts itself to spatial and geometrical modelling. Motion-planning is used mainly in real-time guidance applications and deals with generating a feedback control law to provide torques to manipulator joints and/or drive wheels, in order to track a supplied reference trajectory. The obstacle avoidance is primarily based on a ‘sense and avoid’ philosophy, utilizing the on-board sensors. On the other hand, path-planning finds application mainly in high-level off-line navigation tasks. The output of path-planning exercise is a feasible/optimal trajectory or path through the associated space of possible configurations of the robot and the known obstacles in the working area, from a given initial position to the desired final position of the robot. Genetic algorithms [1] are a powerful tool based on models of natural selection and evolution and facilitate an exhaustive search over large discontinuous spaces (Chapter 1).
Image feature detection using an improved implementation of maximally stable extremal regions for augmented reality applications
Published in International Journal of Image and Data Fusion, 2018
Neetika Gupta, Mukesh Kumar Rohil
Markerless AR, on the other hand, makes use of natural features from the real scene to evaluate the pose and orientation of virtual objects that are to be integrated in the real scene. Keypoint detection in the form of stable image features form a basis for computation in many computer vision tasks like robot navigation, image retrieval, 3D construction, building an AR system, etc. These image features are expected to be highly stable and invariant to affine transformations. Such features are often regions with some distinguished properties from their neighbouring pixels determining them as regions of interest (Lin et al. 2009). These detected regions of interest are combined with a suitable descriptor for attaining more precise affine invariance properties. These descriptors define the region surrounding a point of interest with some distinguishable property that helps in more accurate feature tracking results in a series of image under different affine transformations (Lin et al. 2009, Gauglitz et al. 2011).
RUR53: an unmanned ground vehicle for navigation, recognition, and manipulation
Published in Advanced Robotics, 2021
Nicola Castaman, Elisa Tosello, Morris Antonello, Nicola Bagarello, Silvia Gandin, Marco Carraro, Matteo Munaro, Roberto Bortoletto, Stefano Ghidoni, Emanuele Menegatti, Enrico Pagello
The implemented exploration routine exploits the ROS Navigation Stack: a ROS toolbox which provides robot navigation, collision avoidance, and SLAM (Simultaneous Localization and Mapping). Dynamic Window Approach (DWA) [10] and Timed Elastic Bands (TEB) [11] are considered as local planners, while the GMapping algorithm [12, 13] is used for SLAM. Moreover, way-points can be manually assigned to make the navigation routine compute robot's trajectory while covering these points.
A benchmark of selected algorithmic differentiation tools on some problems in computer vision and machine learning
Published in Optimization Methods and Software, 2018
Filip Srajer, Zuzana Kukelova, Andrew Fitzgibbon
In computer vision, 3D reconstruction is a widely studied problem [3,23]. Given a visual input (e.g. images or video) observing the same scene, the goal is to reconstruct a 3D model of this scene. Even though the creation of 3D models can be a goal on its own, 3D reconstruction is necessary for a number of other applications such as localization [20], robot navigation [8], augmented reality or virtual reality [21].