Autonomous navigation in dynamic environments is hindered by two fundamental challenges: perception instability and uncertainty-optimization mismatch. The former leads to identity switches and unreliable motion estimation, while the latter prevents principled incorporation of motion uncertainty into trajectory optimization. To address these challenges, we propose UCON, an uncertainty-aware navigation algorithm in dynamic environments. For perception instability, we present a point-level historical re-association mechanism that leverages historical point cloud fragments to recover lost targets while maintaining identity continuity. Subsequently, a Kalman filter is employed to provide anisotropic motion state estimation and covariance propagation. To resolve the uncertainty-optimization mismatch, we transform predicted states and their covariances into uncertainty sectors, which are embedded as differentiable cost terms within a trajectory optimization framework. This achieves consistent uncertainty-aware dynamic obstacle avoidance while maintaining smoothness and feasibility. Extensive simulations and real-world experiments demonstrate that, while maintaining high computational efficiency, UCON achieves superior perception stability and robust navigation performance in dynamic environments compared to state-of-the-art methods. The code will be open-sourced to facilitate further research.
Figures & tables
Fig. 1: Comparison between the existing methods and the proposed UCON. (a) Limitations of existing dynamic navigation approaches. (b) The proposed UCON maintains identity continuity via historical re-association and embeds anisotropic motion uncertainty in trajectory optimization.
Fig. 2: Overview of the proposed UCON. UCON integrates stable perception with uncertainty-aware trajectory optimization. Taking LiDAR point clouds and robot pose as inputs, the system first distinguishes between dynamic and static obstacles. The stable perception module initially employs an object-level association mechanism to detect dynamic objects, followed by a point-level historical re-association mechanism to recover lost objects, thereby ensuring perception stability. The uncertainty-aware planning module estimates the motion prediction covariance of dynamic objects and generates uncertainty sectors. During the trajectory optimization stage, these sectors are explicitly converted into differentiable cost functions for dynamic obstacle avoidance, ultimately producing a smooth and safe trajectory.
Fig. 3: Illustration of point-level historical re-association. (a) The perception module needs to identify three dynamic objects and a static wall. (b) Traditional object-level association methods suffer from perception instability when dynamic objects are close to static objects or when two dynamic objects are close to each other. (c) The proposed point-level historical re-association mechanism re-associates points based on historical information. (d) The three dynamic objects are successfully detected, and their identities remain consistent.
Fig. 4: Illustration of uncertainty-aware sectors. Each sector is constructed based on the predicted state and covariance matrix of dynamic obstacles.
Fig. 5: The omnidirectional robot used in real-world experiments. The LiDAR is mounted upside down at the top to fully perceive the robot’s surrounding environment. All computations are performed on the onboard Jetson AGX Orin platform.
Fig. 6: Keyframes from the perception stability benchmark in a real-world environment. (a) Thanks to the proposed point-level historical re-association mechanism, UCON effectively prevents target loss and ensures that the target’s identity remains unchanged, thereby greatly mitigating perception instability. (b) The previous method FAPP relies solely on a single-metric object-level association mechanism, frequently resulting in target loss and frequent identity switching, causing perception instability.
Intent-MPC
FAPP
UCON (ours)
MOTA
66.4%
70.3%
85.2%
Identity Switch
17
16
3
TABLE I: Comparison of Perception Stability
Fig. 7: Comparison of trajectories generated in a simulation environment with 80 dynamic obstacles. (a) Due to the incomplete modeling of stability uncertainty for dynamic obstacles, the trajectories generated by FAPP fail to reserve sufficient safety margins, leading to collisions. (b) The trajectory generated by the learning-based Intent-MPC is not ideal due to its low generalization ability. (c) The trajectory generated by the ablation method is unsafe because it does not consider the motion uncertainty of dynamic obstacles. (d) The complete UCON generates a safe and smooth trajectory.
Scenario
Method
E (m/s 3 )
tp (ms)
η (%)
Low Density Nd=50
Intent-MPC
8.01
40.23
78
FAPP
1.39
2.33
88
UCON (Ablation)
1.28
1.87
84
UCON (Complete)
1.57
2.24
90
Mid Density Nd=80
Intent-MPC
12.70
58.51
62
FAPP
1.65
2.59
78
TABLE II: Comparison of Trajectory Optimization
Fig. 8: Visualization of real-world experiments. The images on the left illustrate the challenges faced by the robot in experimental environments, while the images on the right display the optimized trajectories of the robot. For privacy protection, visible faces are anonymized in the figures.
Visual navigation in unknown environments remains a core challenge in mobile robotics, especially for resource-constrained platforms. Most existing approaches rely on loosely coupled modular pipelines and strong assumptions on perception quality or environmental structure, often resorting to multi-modal sensor suites that increase system complexity and deployment cost. Vision-only navigation offers a lightweight alternative, but its performance degrades severely under motion blur, low texture, and illumination changes, largely because they neglect the tight coupling between commanded motion and perception. While perception-aware methods partially address this issue, they typically optimize individual modules and fail to propagate uncertainty consistently across the navigation stack. In this paper, we present UNSEEN, a unified uncertainty- and perception-aware navigation framework that explicitly couples localization, mapping, and planning using only a front-mounted camera. UNSEEN estimates sparse maps and robot poses with associated uncertainties at 6Hz, and leverages them to plan trajectories that jointly optimize task progress and estimation accuracy in receding-horizon. Simulations and extensive real-world experiments in unknown environments demonstrate the robustness of the proposed approach, with UNSEEN-SLAM reducing absolute translational error by 9.8% and UNSEEN-Plan improving estimation accuracy by up to 45% compared to state-of-the-art methods, while achieving a 100% task success rate.
Tommaso Faraci, Marco Camurri, Daniele Fontanelli +1
Deep reinforcement learning (DRL) finds extensive application in autonomous drone navigation within complex, high-risk environments. However, its practical deployment faces a safety-exploration dilemma: soft penalty mechanisms encourage risky trial-and-error, while most constraint-based methods suffer degraded performance under sensor noise and intent uncertainty. We propose Dynamic-TD3, a physically enhanced framework that enforces strict safety constraints while maintaining maneuverability by modeling navigation as a Constrained Markov Decision Process (CMDP). This framework integrates an Adaptive Trajectory Relational Evolution Mechanism (ATREM) to capture long-range intentions and employs a Physically Aware Gated Kalman Filter (PAG-KF) to mitigate non-stationary observation noise. The resulting state representation drives a dual-criterion policy that balances mission efficiency against hard safety constraints via Lagrangian relaxation. In experiments with aggressive dynamic threats, this approach demonstrates superior collision avoidance performance, reduced energy consumption, and smoother flight trajectories.
We address the problem of reactive motion planning for quadrotors operating in unknown environments with dynamic obstacles. Our approach leverages a 4-dimensional spatio-temporal planner, integrated with vision-based Safe Flight Corridor (SFC) generation and trajectory optimization. Unlike prior methods that rely on map fusion, our framework is mapless, enabling collision avoidance directly from perception while reducing computational overhead. Dynamic obstacles are detected and tracked using a vision-based object segmentation and tracking pipeline, allowing robust classification of static versus dynamic elements in the scene. To further enhance robustness, we introduce a backup planning module that reactively avoids dynamic obstacles when no direct path to the goal is available, mitigating the risk of collisions during deadlock situations. We validate our method extensively in both simulation and real-world hardware experiments, and benchmark it against state-of-the-art approaches, showing significant advantages for reactive UAV navigation in dynamic, unknown environments.
Astik Srivastava, Thomas J Chackenkulam, Bitla Bhanu Teja +2
Robotics Research Center, IIIT Hyderabad, India · IIT-BHU, India