The joint optimization of map management and map feature to measurement association, together with the trajectory and map states, within a single, unified, Bayesian, feature-based, simultaneous localization and mappin...
详细信息
The joint optimization of map management and map feature to measurement association, together with the trajectory and map states, within a single, unified, Bayesian, feature-based, simultaneous localization and mapping (SLAM) solution is addressed in this article. Remarkable progress in feature-based SLAM has been made in which, given data association, the SLAM problem can be solved by use of nonlinear least squares solvers, often referred to as the SLAM back-end. These methods rely on external methods to solve both the data association and map management problems, which are collectively incorporated into the SLAM front-end. SLAM convergence failures are common when these front-end routines fail, particularly when feature detection uncertainty increases. Therefore, this article introduces Joint, Vector-Set SLAM (JVS-SLAM), utilizing Bayes theorem to solve feature to measurement association, map management, and SLAM itself jointly, thus combining the SLAM back and front ends. Results will demonstrate equivalent or superior SLAM performance to state-of-the-art solutions, under varying odometry, spatial and detection measurement uncertainties, without reliance on data association decisions. Results are based on both simulations and the challenging EuRoC data set, in which a drone undergoing high accelerations, equipped with a stereo camera, performs SLAM. Since JVS-SLAM jointly provides a solution to the map feature to measurement association problem, its computational complexity is comparable with multi-hypothesis based solutions. Parallels between state-of-the-art map management and feature to measurement association methods and the detection statistics used within JVS-SLAM will be examined, with a view to reducing its complexity in the future.
A tensegrity-based system is a promising approach for dynamic exploration of uneven, unpredictable, and confined environments. However, implementing such systems presents challenges in state recognition. In this study...
详细信息
A tensegrity-based system is a promising approach for dynamic exploration of uneven, unpredictable, and confined environments. However, implementing such systems presents challenges in state recognition. In this study, we introduce a 6-strut tensegrity structure integrated with 24 multimodal strain sensors, employing a deep learning model to achieve smart tensegrity. By using conductive flexible tendons and leveraging a long short-term memory (LSTM) model, the system accomplishes self-shape reconstruction without the need for external sensors. The sensors operate in two modes, and we applied both a curve fitting model and an LSTM model to establish the relationship between length change and resistance change in the sensors. Our key findings demonstrate that the intelligent tensegrity system can accurately self-detect and adapt its shape. Furthermore, a human pressing process allows users to monitor and understand the tensegrity's shape changes based on the integrated models. This intelligent tensegrity-based system with self-sensing tendons showcases significant potential for future exploration, making it a versatile tool for real-world applications.
Intelligent virtual agents are used to accomplish complex multi-modal tasks such as human instruction comprehension in mixed-reality environments by increasingly adopting richer, energy-intensive sensors and processin...
详细信息
Intelligent virtual agents are used to accomplish complex multi-modal tasks such as human instruction comprehension in mixed-reality environments by increasingly adopting richer, energy-intensive sensors and processing pipelines. In such applications, the context for activating sensors and processing blocks required to accomplish a given task instance is usually manifested via multiple sensing modes. based on this observation, we introduce a novel Commit-and-Switch (CAS) paradigm that simultaneously seeks to reduce both sensing and processing energy. In CAS, we first commit to a low-energy computational pipeline with a subset of available sensors. Then, the task context estimated by this pipeline is used to optionally switch to another energy-intensive DNN pipeline and activate additional sensors. We demonstrate how CAS's paradigm of interweaving DNN computation and sensor triggering can be instantiated principally by constructing multi-head DNN models and jointly optimizing the accuracy and sensing costs associated with different heads. We exemplify CAS via the development of the RealGIN-MH model for multi-modal target acquisition tasks, a core enabler of immersive human-agent interaction. RealGIN-MH achieves 12.9x reduction in energy overheads, while outperforming baseline dynamic model optimization approaches.
The application of artificial intelligence (AI) and robotics in extreme environments, is crucial for addressing complex challenges and performing high risk tasks. We highlight the importance of multi-modal sensor redu...
详细信息
ISBN:
(纸本)9798350364200;9798350364194
The application of artificial intelligence (AI) and robotics in extreme environments, is crucial for addressing complex challenges and performing high risk tasks. We highlight the importance of multi-modal sensor redundancy to ensure system reliability and accuracy despite sensor failures caused by harsh environmental conditions. We propose design considerations for sensors in extreme environments, emphasizing both the hardware and software design. One method is the non-contact heart rate and temperature monitoring using RGB visible and infrared cameras. This method addresses the limitations of traditional visible light sensors under complex illumination conditions, enhancing data reliability through advanced data fusion techniques. Furthermore, we propose a panoramic sensor lens design with a 270-degree view for comprehensive environmental perception, reducing mechanical vulnerabilities. These designs demonstrate the effectiveness of combining infrared and visible light sensors for improved environmental perception and physiological monitoring.
This article solves the planar navigation problem by recourse to an online reactive scheme that exploits recent advances in simultaneous localization and mapping (SLAM) and visual object recognition to recast prior ge...
详细信息
This article solves the planar navigation problem by recourse to an online reactive scheme that exploits recent advances in simultaneous localization and mapping (SLAM) and visual object recognition to recast prior geometric knowledge in terms of an offline catalog of familiar objects. The resulting vector field planner guarantees convergence to an arbitrarily specified goal, avoiding collisions along the way with fixed but arbitrarily placed instances from the catalog as well as completely unknown fixed obstacles so long as they are strongly convex and well separated. We illustrate the generic robustness properties of such deterministic reactive planners as well as the relatively modest computational cost of this algorithm by supplementing an extensive numerical study with physical implementation on both a wheeled and legged platform in different settings.
We propose a novel belief space planning technique for continuous dynamics by viewing the belief system as a hybrid dynamical system with time-driven switching. Our approach is based on the perturbation theory of diff...
详细信息
We propose a novel belief space planning technique for continuous dynamics by viewing the belief system as a hybrid dynamical system with time-driven switching. Our approach is based on the perturbation theory of differential equations and extends sequential action control to stochastic dynamics. The resulting algorithm, which we name SACBP, does not require discretization of spaces or time and synthesizes control signals in near real-time. SACBP is an anytime algorithm that can handle general parametric Bayesian filters under certain assumptions. We demonstrate the effectiveness of our approach in an active sensing scenario and a model-based Bayesian reinforcement learning problem. In these challenging problems, we show that the algorithm significantly outperforms other existing solution techniques including approximate dynamic programming and local trajectory optimization.
暂无评论