Learning Feasibility for Task and Motion Planning in Tabletop Environments.

Robots used for inspection, package deliveries, moving of goods, and other logistics operations are often required to visit certain locations within specified time bounds. This gives rise to a challenging problem as it requires not only planning collision-free and dynamically feasible motions but also reasoning temporally about when and where the robot should be. While significant progress has been made in integrating task and motion planning, there are still no effective approaches for multigoal motion planning when both dynamics and time windows must be satisfied. To effectively solve this challenging problem, this paper develops an approach that couples temporal planning over a discrete abstraction with sampling-based motion planning over the continuous state space of feasible motions. The discrete abstraction is obtained by imposing a roadmap that captures the connectivity of the free space. At each iteration of a core loop, the approach first invokes the temporal planner to find a solution over the roadmap abstraction. In a second step, the approach uses sampling to expand a motion tree along the regions associated with the discrete solution. Experiments are conducted with second-order ground and aerial vehicle models operating in complex environments. Results demonstrate the efficiency and scalability of the approach as we increase the number of goals and the difficulty of satisfying the time bounds.

This study presents a hybrid robotics automation system that optimizes real-time navigation, obstacle avoidance, and environmental mapping under difficult situations by combining AutoNav, LIDAR-based SLAM, and DenseNet with Leaky ReLU for robust motion planning. The main goals are to use AutoNav to enable autonomous navigation with optimal path planning, LIDAR-based SLAM to accomplish precise mapping and localization, and DenseNet and Leaky ReLU to improve motion planning and obstacle avoidance. Through the integration of various technologies, the study suggests an effective and flexible system for robotic operations in dynamic situations in real time. In order to provide the best trajectory planning and accurate obstacle recognition, the system constantly modifies motion planning based on real-time sensor data. Performance results show that the system outperforms individual parts, with a 95% gradient efficiency, a localization error of only 0.045 meters, and a pathfinding accuracy of 97.5%, the system's performance results illustrate its superiority over individual parts and highlight its real-time adaptability and optimization accuracy. For the future of robotics, this hybrid system offers breakthroughs in real-time decision-making and nimble autonomous robotic navigation.

Motion Planning (MP) is a critical challenge in robotics, especially pertinent with the burgeoning interest in embodied artificial intelligence. Traditional MP methods often struggle with high-dimensional complexities. Recently neural motion planners, particularly physics-informed neural planners based on the Eikonal equation, have been proposed to overcome the curse of dimensionality. However, these methods perform poorly in complex scenarios with shaped robots due to multiple solutions inherent in the Eikonal equation. To address these issues, this paper presents PC-Planner, a novel physics-constrained self-supervised learning framework for robot motion planning with various shapes in complex environments. To this end, we propose several physical constraints, including monotonic and optimal constraints, to stabilize the training process of the neural network with the Eikonal equation. Additionally, we introduce a novel shape-aware distance field that considers the robot's shape for efficient collision checking and Ground Truth (GT) speed computation. This field reduces the computational intensity, and facilitates adaptive motion planning at test time. Experiments in diverse scenarios with different robots demonstrate the superiority of the proposed method in efficiency and robustness for robot motion planning, particularly in complex environments.

This article is concerned with methods of computing a path in 3-DOF space that describes the desired translational motion of an aerial vehicle. Trajectory planning is an optimization problem which generates an optimal trajectory between two configurations in the state space, considering a given performance index (time, energy or distance). Its feasibility depends on the choice of the optimization method, the performance index and a number of constraints from various nature, the latter depending essentially on the vehicle itself (architecture, dynamics and actuation modes) and the environment in which the vehicle moves (endurance, airspeed, altitude, landing and takeoff modes, etc.) [3, 5, 12, 13]. The classical differential geometry curve theory is a study of 3D space curves with orthogonal coordinate systems attached to moving points on the space curve [1, 2, 4, 11]. The atmosphere is considered to be an isotropic and homogeneous medium, i.e. when there is no wind and the air density is constant with altitude. Classically, in motion planning and generation, methods such as continuous optimization and discrete search are sought. In [9] randomized motion planning algorithm by employing obstacle free guidance system as local planners in a probabilistic road-map framework are presented. In [13], path planning of autonomous fixed wing aircraft is based on a learning real-time A* search algorithm, considering only the motion on a horizontal plane. A family of trim trajectories in level flight is used in all these references to construct paths. In [7], plans are described as the concatenation of a number of well defined motion primitives selected from a finite library.

Earlier work has shown that reusing experience from prior motion planning\nproblems can improve the efficiency of similar, future motion planning queries.\nHowever, for robots with many degrees-of-freedom, these methods exhibit poor\ngeneralization across different environments and often require large datasets\nthat are impractical to gather. We present SPARK and FLAME , two\nexperience-based frameworks for sampling-based planning applicable to complex\nmanipulators in 3 D environments. Both combine samplers associated with\nfeatures from a workspace decomposition into a global biased sampling\ndistribution. SPARK decomposes the environment based on exact geometry while\nFLAME is more general, and uses an octree-based decomposition obtained from\nsensor data. We demonstrate the effectiveness of SPARK and FLAME on a Fetch\nrobot tasked with challenging pick-and-place manipulation problems. Our\napproaches can be trained incrementally and significantly improve performance\nwith only a handful of examples, generalizing better over diverse tasks and\nenvironments as compared to prior approaches.\n

We propose the combination of techniques that solve multiple queries for motion planning problems with single query planners in a motion planning framework that can be efficiently parallelized. In multiple query motion planning, a data structure is built during a preprocessing phase in order to quickly respond to on-line queries. Alternatively, in single query planning, there is no preprocessing phase and all computations occur during query resolution. This paper shows how to effectively combine a powerful sample-based method primarily designed for multiple query planning (the Probabilistic Roadmap Method - PRM) with sample-based tree methods that were primarily designed for single query planning (such as Expansive Space Trees, Rapidly Exploring Random Trees, and others). Our planner, which we call the Probabilistic Roadmap of Trees (PRT), uses a tree algorithm as a subroutine for PRM. The nodes of the PRM roadmap are now trees. We take advantage of the very powerful sampling schemes of recent tree planners to populate our roadmaps. The combined sampling scheme is in the spirit of the non-uniform sampling and refinement techniques employed in earlier work on PRM. PRT not only achieves a smooth spectrum between multiple query and single query planning but it combines advantages of both. We present experiments which show that PRT is capable of solving problems that cannot be addressed efficiently with PRM or single-query planners. A key advantage of PRT is that it is significantly more decoupled than PRM and sample-based tree planners. Using this property, we designed and implemented a parallel version of PRT. Our experiments show that PRT distributes well and can easily solve high dimensional problems that exhaust resources available to single machines.

Soft robots, owing to their inherent compliance and adaptability, are particularly well-suited for on-orbit applications. However, motion planning for such robots in space environments poses considerable challenges, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e.g.</i>, navigating through cluttered orbital debris, accounting for microgravity-induced dynamic behaviors during obstacle avoidance. This paper presents a motion planning framework tailored for spacecraft-mounted soft manipulators (SMSMs). A generalized practical finite-time stable optimization algorithm is proposed to enable efficient and robust planning. By leveraging strict convexification and discretization, the method achieves rapid convergence while accommodating the complex dynamics and morphological characteristics of SMSMs. To further enhance autonomy, a generative adversarial network integrated with long short-term memory (GAN-LSTM) is developed to learn intricate state-control mappings. This deep learning architecture captures temporal dependencies to improve the quality of control policies and reduce reliance on manual intervention or re-planning. The effectiveness of all proposed methods is validated through simulations.

A novel model predictive artificial potential field based ship motion planning method considering COLREGs for complex encounter scenarios

This article addresses the localization problem in robotic autonomous luggage trolley collection at airports and provides a systematic evaluation of different methods to solve it. The robotic autonomous luggage trolley collection is a complex system that involves object detection, localization, motion planning and control, manipulation, etc. Among these components, effective localization is essential for the robot to employ subsequent motion planning and end-effector manipulation because it can provide a correct goal position. This article explores four popular and representative localization methods for object localization in luggage trolley collection: radio frequency identification (RFID), Keypoints, ultrawideband (UWB), and Reflectors. A qualitative evaluation framework is constructed to assess performance, encompassing Localization Accuracy, Mobile Power Supplies, Coverage Area, Cost, and Scalability. Furthermore, a series of quantitative experiments concerning Localization Accuracy and Success Rate have been conducted on a real-world robotic autonomous luggage trolley collection system. The performance of various localization methods is further analyzed based on experimental results, indicating that the Keypoints method is optimally suited for indoor environments to facilitate luggage trolley collection. Significantly, these experiment results provide a valuable reference point, extending the application of indoor localization methods across diverse scenarios. A website about this work is available at https://sites.google.com/view/localization-evaluation/.

Obstacle avoidance is essential for ensuring the safety of autonomous vehicles. Accurate perception and motion planning are crucial to enabling vehicles to navigate complex environments while avoiding collisions. In this paper, we propose an efficient obstacle avoidance pipeline that leverages a camera-only perception module and a Frenet-Pure Pursuit-based planning strategy. By integrating advancements in computer vision, the system utilizes YOLOv11 for object detection and state-of-the-art monocular depth estimation models, such as Depth Anything V2, to estimate object distances. A comparative analysis of these models provides valuable insights into their accuracy, efficiency, and robustness in real-world conditions. The system is evaluated in diverse scenarios on a university campus, demonstrating its effectiveness in handling various obstacles and enhancing autonomous navigation. The video presenting the results of the obstacle avoidance experiments is available at: https://www.youtube.com/watch?v=FoXiO5S_tA8

Computation of the volume of space required for a robot to execute a sweeping motion from a start to a goal has long been identified as a critical primitive operation in both task and motion planning. However, swept volume computation is particularly challenging for multi-link robots with geometric complexity, e.g., manipulators, due to the non-linear geometry. While earlier work has shown that deep neural networks can approximate the swept volume quantity, a useful parameter in sampling-based planning, general network structures do not lend themselves to outputting geometries. In this paper we train and evaluate the learning of a deep neural network that predicts the swept volume geometry from pairs of robot configurations and outputs discretized voxel grids. We perform this training on a variety of robots from 6 to 16 degrees of freedom. We show that most errors in the prediction of the geometry lie within a distance of 3 voxels from the surface of the true geometry and it is possible to adjust the rates of different error types using a heuristic approach. We also show it is possible to train these networks at varying resolutions by training networks with up to 4x smaller grid resolution with errors remaining close to the boundary of the true swept volume geometry surface.

Advanced upper-limb prostheses capable of actuating multiple degrees of freedom (DOF) are now commercially available. Pattern recognition based algorithms that use surface electromyography (EMG) signals measured from residual muscles show great promise as multi-DOF controllers. Unfortunately, current pattern recognition systems are limited to sequential control of each DOF. This study introduces a hierarchy of linear discriminant analysis (LDA) classifiers arranged to provide simultaneous DOF control. This approach and two other simultaneous control strategies were evaluated using healthy subjects controlling up to four DOFs, where any two DOFs could be controlled simultaneously. The new hierarchical approach was the most promising with classification errors at or below 15% on average for discrete and combined motions. The classification performance was significantly better (p < 0.05) than using a single LDA classifier trained to recognize both discrete and combined motions or classifying each DOF using a set of parallel classifiers. The high accuracy of the hierarchical approach suggests that pattern recognition techniques can be extended to permit simultaneous control, potentially allowing amputees to produce more fluid, life-like movements, ultimately increasing their quality of life.

Reverse k-rank queries, for a given query item, extract the top-k users whose ranking of the item (based on individual user’s preferences) is best among all the users. Such queries have recently received significant interest in diverse applications like market analysis, product placement, sales and e-commerce. Current approaches employ efficient high-dimensional data indexing techniques to prune input data points for improving query run-time. However, they fail to provide practical run-time characteristics for online and streaming scenarios, typical in such applications. This paper proposes the RADAR algorithm to efficiently compute, in real-time, approximate reverse k-rank queries. RADAR sorts the input data on each dimension (i.e., item aspects), and utilizes the ranking of a query in each of the dimensions to approximate the final ranking of the query by users based on their preferences. Empirical evaluations on real datasets demonstrates upto 50\(\times \) run-time improvements over existing approaches, with a high accuracy of around \(90\%\). Further, experiments on synthetic datasets showcase the scalability and efficacy of our algorithm for large scale and high-dimensional datasets.

Humanoid robots are expected to play a big role at distress sites and disaster sites. There is a variety of multi-contact locomotion forms other than bipedal walking such as crawling through tightly, getting on the rubble by using its knees and elbows, or jumping in and rolling over the obstacles. If such multi-contact locomotion forms can be achieved, robots can reach environments that are currently unreachable, and be able to conduct tasks required at the environments. To achieve this, it is required for robots to bring various parts of its body into contact with the environment like a human. However, it is difficult for parts without 6-axis force sensors to achieve the target force while adapting to the environment against impact force. It is also difficult to measure contact wrenches without 6-axis force sensors. In this paper, by allowing the error of the contact state, we propose online system for realizing dynamic motion which impact force occurs on the parts of the whole body by contact to the environment. In the proposed system, we applied the current-based torque control for joints to make the whole body parts of the robot adapt to the environment, and we modified motion in real time to stabilize zmp by estimating contact wrenches at the contact positions where force sensors are not mounted. In addition, at the motion planning, we generated more feasible motions for a robot applying torque control by using evolutionary computation which advances the search with the behavior of torque control. We demonstrate that the proposed system is effective by showing experimental results of sitting posture locomotion using a JAXON robot in which impact force occur on the back of the thighs which have no force sensors.

Motion planning for unmanned surface vehicle based on a maneuverability mathematical model