This paper addresses autonomous intervention with an underwater vehicle--manipulator system (UVMS) in confined, cluttered, and partially known environments, where poor maneuverability, narrow passages, and uncertain execution may cause the robot to enter unrecoverable regions. We propose MANTA, a three-layer hierarchical planning-and-control framework that couples passage accessibility, manipulation feasibility, and closed-loop execution. The first layer performs global connectivity reasoning in a conservative reduced base space to extract traversable corridor candidates toward the task region. The second layer refines each candidate corridor by jointly optimizing the continuous base motion and arm trajectory, producing a collision-free base--arm trajectory. The third layer learns a reach-and-hold base policy using Gaussian-process model-based reinforcement learning (MBRL) through MC-PILCO, enabling trajectory tracking and station keeping at the planned manipulation state. During execution, the framework monitors map updates and can trigger recovery and route repair when the active passage becomes infeasible. MANTA is evaluated in confined UVMS planning and closed-loop tracking experiments. Across 120 matched planning queries, it achieves higher task success than full-state sampling-based baselines while producing larger clearance margins and lower arm motion. The learned MC-PILCO policy further reduces position and yaw tracking errors on both training and unseen tube-like references. These results show MANTA as a structured and data-efficient framework for safe autonomous underwater intervention in caves, tubes, and cluttered subsea structures.
Most UAV-based transmission line inspection systems still use preset waypoints and treat task planning as separate from motion control. Single-layer methods, such as waypoint-based, learning-based or MPC-based methods, generally handle planning or control independently; none coordinate task reallocation with trajectory adjustment in the event of wind disturbances or sensor faults. We propose a hierarchical framework that connects task-level plans with motion-level control via intelligent agents operating at two different timescales. A high-level strategic agent divides the task into multiple areas. Several low-level execution agents plan local trajectories and coordinate via a distributed consensus protocol. A conditional value-at-risk (CVaR)-aware model predictive controller is employed for motion control to address wind disturbances, obstacle constraints and actuator limits; fault detection activates an adaptive replanning mechanism to maintain operation under degraded sensor conditions or communication loss. In simulation on a 15-km corridor with three UAVs, the framework achieved 15.7% higher inspection coverage than conventional waypoint-based methods, 17.5% lower energy consumption, and a tracking error of less than 0.35 m at crosswinds of up to 12 m/s. Multi-UAV coordination was validated in simulation; a single-UAV field experiment on a 110 kV line segment validated autonomous inspection with 93.3% coverage and approximately 0.5 m positioning accuracy.
Huanyu Cheng, Yingcheng Gu, Mengting Xi et al.· IEEE Access· 0 citations
Robotic manipulators operating in cluttered environments require collision-free trajectories that remain executable under kinematic and dynamic constraints. This paper proposes a reinforcement learning (RL)-guided multi-objective trajectory planning framework, termed RL-MOP-HNE, for a 6-DOF UR5 manipulator. The planning model simultaneously minimizes path length, energy consumption, and execution time while satisfying collision-avoidance, kinematic, and dynamic constraints. A tabular SARSA agent is embedded into the evolutionary search to adaptively select search behaviours according to the current optimization state. To improve the balance between exploration and exploitation, a Gaussian-perturbation adaptive hybrid crossover operator is integrated with a hierarchical neighborhood evolution (HNE) strategy, enabling progressive population refinement throughout the search process. The proposed method is evaluated in three representative environments with increasing planning complexity, including single-obstacle, narrow three-obstacle, and irregular five-obstacle scenarios, and is compared with MOEA/D, MOPSO, MSCLPSO, NSGA-II, and RL-NSGA-II. Experimental results show that RL-MOP-HNE generates feasible trajectories in all test cases and achieves the lowest dynamic-stability-prioritized composite scores among the compared algorithms. The planned trajectories exhibit smoother joint motion and lower velocity fluctuations, although these improvements are generally accompanied by longer execution times. Complementary analyses, including time scaling, manipulability, clearance evaluation, statistical significance tests, and ablation studies, further explain the performance characteristics of the proposed framework and quantify the contribution of its key components. The proposed framework is therefore well suited to robotic applications where motion stability and dynamic executability are of greater importance than minimum-time operation.
Zhen-Long Zhao, Shu-Tao Hao, Bi-Hao Jin et al.· Scientific Reports· 0 citations
This paper presents an observation-only autonomy framework for Unmanned Underwater Vehicles (UUVs) navigation in dynamic underwater environments that integrates persistent occupancy mapping, global clearance-aware planning, and risk-aware local control. The proposed pipeline constructs occupancy maps solely from onboard sonar and depth image observations, adapts a clearance-constrained global planner (GP) to provide long-horizon structure, and integrates a reinforcement learning (RL) policy to handle short-range tracking and reactive avoidance. To further support decision-making under partial observability, the system learns a compact latent state representation from onboard sensor data, encoding environmental structure, obstacle dynamics, and uncertainty. Behavior tree (BT) distillation with staged supervision is introduced to improve safety and training stability, while an uncertainty-calibrated distillation mechanism reweights teacher guidance using online latent-model uncertainty, emphasizing uncertain regimes during learning, with time-to-collision (TTC) and clearance cues remaining explicit in planning and local policy features. To demonstrate the efficacy of the framework, a reproducible multi-seed evaluation protocol is established in high-fidelity GPU-accelerated simulation using NVIDIA Isaac Sim, and performance is benchmarked against BT-only and standard RL baselines. The results obtained demonstrate improved robustness and safety under dynamic conditions, thus providing a general pipeline with a unified hybrid planning learning architecture and a reproducible methodology for robust UUV autonomy under partial observability.
Low-altitude wireless networks (LAWNs) are emerging as a foundational infrastructure for the low-altitude economy, necessitating unmanned aerial vehicles (UAVs) to operate as embodied wireless agents that can navigate complex air-ground environments. This paper investigates the critical challenge of maintaining reliable and covert information transmission for such agents under stringent mobility and detection constraints. To address this, we propose an embodied agentic framework that synergizes safety-critical path planning with cognitive predictive beamforming through a hierarchical execution-reasoning loop. Specifically, the execution layer utilizes model predictive control (MPC) to solve a constrained optimization problem, linearizing obstacle-avoidance requirements into tractable dual-variable inequalities to ensure collision-free navigation. This generates a deterministic sequence of future coordinates that serves as a spatial intent prior to reduce environmental uncertainty. Subsequently, the cognitive reasoning layer leverages a large language model (LLM) to integrate these motion priors with sensing-assisted historical channel state information and warden locations through a cross-modal attention mechanism. By capturing the intricate dependencies between the agent kinematic states and the non-stationary wireless channel, the LLM proactively optimizes predictive beamforming to satisfy covertness requirements against mobile wardens. Simulation results validate the efficacy of this joint design, demonstrating that the UAV successfully tracks reference trajectories while maintaining safety margins in obstacle-cluttered environments. Furthermore, the proposed framework achieves superior covertness performance compared to conventional benchmarks.
In urban search and rescue, articulated tracked robots (ATRs) must traverse structured but contact-rich environments such as stairwells and cluttered building interiors. Reliable autonomy remains challenging because robot-terrain interaction (RTI) is hybrid and discontinuous, and effective flipper-track coordination is difficult to model analytically. We present ASTRIL-MPC, a language-guided neural kinematics model predictive control (MPC) framework for autonomous traversal. A learned kinematics model predicts short-horizon task-state increments from a height sequence and recent trajectories; NMPC plans with multi-objective costs and strict feasibility constraints; and a large language model (LLM) proposes bounded updates to selected weights and bounds through a safety-checked interface with range clipping, rate limiting, and consistency checks. The compiled predictor enables a full control cycle within 100 ms. Across three traversal tasks and a multi-height generalization setting, ASTRIL-MPC improves an aggregate traversal-quality score by up to 71% over a non-adaptive NMPC and by 67% over a PPO baseline, while eliminating measurable collision impacts during descent. These results indicate that combining terrain-conditioned neural kinematics, optimization-based planning, and language-guided adaptation yields data-efficient and robust autonomy for articulated tracked robots. Real-robot trials over four indoor obstacles further demonstrate transfer to contact-rich physical traversal.
Zhe Gan, Yan-Bo Chen, Li-Rong Che et al.· 0 citations
Mobile robots operating in human-populated environments must navigate complex, multi-room spaces while ensuring safety, i.e., generating collision-free motion. In this study, we present a sensor-based model predictive control (MPC) scheme designed for safe crowd navigation in such non-convex environments. The proposed framework decomposes the free space into a set of overlapping convex regions to construct a topological graph, enabling a high-level planner to compute optimal sequences of traversable areas. To effectively perceive the crowd, the system employs a robust perception pipeline that fuses 2D LiDAR data with semantic information from an RGB-D camera, utilizing Kalman filters (KFs) to estimate and predict human motion. These predictions are integrated into an MPC controller which generates robot commands by enforcing safety through discrete-time control barrier function (DT-CBF), ensuring that the robot avoids collisions while remaining within navigable regions. The approach is validated through high-fidelity simulations and real-world experiments using the TIAGo mobile manipulator. The results demonstrate that integrating vision-based semantic data with geometric constraints significantly improves collision avoidance and success rates in cluttered, multi-room scenarios.
Giovanbattista Gravina, Francesco D'Orazio, Michele Cipriano et al.· Frontiers in Robotics and AI· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
