For the indoor navigation of service robots, human–robot interaction and adapting to the environment still need to be strengthened, including determining the navigation goal socially, improving the success rate of passing doors, and optimizing the path planning efficiency. This paper proposes an indoor navigation system based on object semantic grid and topological map, to optimize the above problems. First, natural language is used as a human–robot interaction form, from which the target room, object, and spatial relationship can be extracted by using speech recognition and word segmentation. Then, the robot selects the goal point from the target space by object affordance theory. To improve the navigation success rate and safety, we generate auxiliary navigation points on both sides of the door to correct the robot trajectory. Furthermore, based on the topological map and auxiliary navigation points, the global path is segmented into each topological area. The path planning algorithm is carried on respectively in every room, which significantly improves the navigation efficiency. This system has demonstrated to support autonomous navigation based on language interaction and significantly improve the safety, efficiency, and robustness of indoor robot navigation. Our system has been successfully tested in real domestic environments.

Active mapping is an important technique for mobile robots to autonomously explore and recognize indoor environments. View planning, as the core of active mapping, determines the quality of the map and the efficiency of exploration. However, most current view-planning methods focus on low-level geometric information like point clouds and neglect the indoor objects that are important for human–robot interaction. We propose a novel View-Planning method for indoor active Sparse Object Mapping (VP-SOM). VP-SOM takes into account for the first time the properties of object clusters in the coexisting human–robot environment. We categorized the views into global views and local views based on the object cluster, to balance the efficiency of exploration and the mapping accuracy. We developed a new view-evaluation function based on objects’ information abundance and observation continuity, to select the Next-Best View (NBV). Especially for calculating the uncertainty of the sparse object model, we built the object surface occupancy probability map. Our experimental results demonstrated that our view-planning method can explore the indoor environments and build object maps more accurately, efficiently, and robustly.

As SLAM technology has evolved from the geometric level to the object level, Active SLAM (ASLAM) has also adopted a new goal: improving the ability to observe objects. However, current ASLAM mostly focuses on low-dimensional environmental features such as points, lines, and planes, while ignores the impact of motion process on SLAM. This paper proposes a new observation pose optimization method based on the ellipsoid model and the camera field of view to enhance the robot’s ability to observe objects. We integrate view planning and motion planning into a unified observation pose optimization module and construct an optimized factor graph based on probabilistic inference. Our method ensures the observation poses are globally optimal and can directly generate robot control variables. Leveraging the camera’s field of view and the ellipsoid model abstracted from object SDF model, we introduce three key factors into the factor graph: object completeness observation factor, self-observation prevention factor, camera motion smoothness factor. Finally, we develop the object ASLAM system with our observation pose optimization method, and evaluate it in multiple simulation environments. Experimental results demonstrate that our method significantly improves object modeling accuracy, mapping efficiency, and localization precision. The code for this work are open-sourced at https://github.com/TINY-KE/OPO_ASLAM.git.

Line-Of-Sight (LOS) check plays a crucial role in collision avoidance and time comsuming, particularly in scenarios involving large-scale maps with sparse obstacles, as it necessitates a grid-by-grid state check. Specifically, LOS check consumes more than half of the computational time in any-angle path planning algorithms, such as Theta*, Visibility Graph, and RRT. To address this issue, we propose an efficient LOS checker for maps of arbitrary dimensions with sparse obstacles. Our approach involves a two-step process. Firstly, we partition the passable space into blocks until there is no vacancy for a minimum-sized block. When the adapted Bresenham algorithm reaches a surface of a block, it bypasses grid-by-grid traversal within the block and directly jumps to the opposing surface. This method significantly reduces the number of grids examined, resulting in higher efficiency compared to traditional LOS checks. We refer to our approach as Jump Over Block (JOB). To demonstrate the advantages of JOB, we compare its performance against traditional LOS checks using a widely recognized public dataset. The results indicate that JOB incurs only 1/6 to 1/5 of the computational cost associated with raw LOS checks, making it a valuable tool for both researchers and practitioners in the field. In order to facilitate further research within the community, we have made the source code of the proposed algorithm publicly available. We anticipate that this framework will contribute to the development of more efficient path planning algorithms and expedite various aspects of robotics that involve collision checks.

The main challenges in object-level SLAM lie in the representation of object geometric properties (position, rotation, scale) and surface shapes, as well as how to efficiently apply such object representation to accurate online camera tracking. Previous methods have under-exploited scene structural priors, limiting the object reconstruction quality and accuracy of localization. In this study, we propose QDSP-SLAM as a novel object-level SLAM system designed for static indoor scenes, which reconstructs objects using both dual quadrics and deep implicit shapes. We introduce the Manhattan world assumption at the object level to establish geometric constraints between dual quadrics and Manhattan planes, thereby significantly improving the accuracy of object geometric properties. Furthermore, we propose a deep implicit shape representation method integrated with dual quadrics to reconstruct object surface shapes, which leverages both category-level shape priors and scene structural priors. We redesigned the bundle-adjustment process specifically for a new object model that integrates dual quadrics and implicit shapes. We validated our approach in both simulated and real indoor scenes, demonstrating significant improvements in both modeling quality and localization accuracy compared with previous methods. The code for this study is available at https://github.com/TINY-KE/QDSP-SLAM.git.