3 May 2017 in Optics Communications
19 July 2023 in IEEE Robotics and Automation Letters
Existing RRT-based exploration methods often suffer from interruptions in the exploration process due to the inability to detect all frontiers of the drivable area in the mapping map. These methods cannot detect all frontiers because the RRT expansion is disturbed by factors such as RRT preset parameters, sliding window constraints, complex external environment, etc, and thus cannot completely cover the drivable area within a limited time. We address this problem by redefining exploration frontiers, designing a novel exploration gain, and constructing minimum RRT search spaces. Our method is evaluated against the existing state-of-the-art RRT-based methods in simulated benchmarks and outdoor environments. The results show that our method is more robust to the above factors while reducing computational cost. Our method is made open source to benefit the community.
30 November 2021 in IEEE Sensors Journal
In this work, a novel paradigm is proposed to build a hybrid RGB/FIR camera, which outputs pixel-wisely aligned and temporally synchronized RGB and FIR frames simultaneously. Compared with the existing schemes, we innovate in several ways to make it compact, practical, and extendable for real-world applications. First, we propose a new structure layout design of the hybrid camera, and we also design a removable adjuster to detach the hybrid camera from the base platform. For RGB and FIR image registration, we have achieved a pixel-wise spatial registration of the FIR and RGB frames by a two-step alignment, where we propose a ghost-effect removal scheme to remove the ghost effect from the rolling-shutter-based FIR camera. Finally, multiple hybrid cameras are extended to a hybrid camera array, through which we can obtain wide-view spatially aligned FIR, RGB and disparity images simultaneously. The experimental results demonstrate the accuracy of the hybrid camera and the effectiveness of the designed stereo wide-view hybrid camera system.
22 September 2026 in SSRN Electronic Journal
Robotic strawberry harvesting remains challenging under partial occlusion, where leaves obscure fruit geometry and makes grasp decisions based on a single shape estimate unreliable. From the same partial observation, multiple 3D completions may be plausible, and a grasp that appears feasible on one completion may fail on another. Existing uncertainty-aware grasping methods typically estimate uncertainty for pose, shape, or individual grasp candidates, but do not explicitly aggregate grasp feasibility across multiple completed object hypotheses to support an object-level attempt-or-abstain decision. This paper presents UNCLE-Grasp, a task-adapted framework that integrates learned shape completion, dropout-based variability estimation, physically grounded grasp evaluation, and risk-aware target-level grasp decisions for leaf-occluded strawberries. Monte Carlo dropout is used to generate multiple completion samples, and for each completion the retained grasp candidates are combined into a single wrench space from which a completion-level force-closure score is computed. Variability of this score across plausible strawberry completions, rather than geometric point variability alone, is then used to quantify target-level grasp uncertainty. A conservative lower confidence bound (LCB) then determines whether the system should execute a grasp or abstain from the target. The framework is evaluated in simulation and on a physical robot under increasing levels of synthetic and real leaf occlusion. At the highest simulated occlusion, UNCLE-Grasp improves success over attempted grasps from 0.780 for the strongest completed baseline to 0.870, while achieving a similar attempt rate of 0.860 versus 0.880 and increasing overall success from 0.680 to 0.740. On the physical robot, it reaches 0.800 success over attempted grasps at the highest synthetic occlusion of approximately 87\%, compared with 0.483 for the strongest completed baseline, although this improvement is accompanied by a lower attempt rate. These results show that the main benefit of UNCLE-Grasp is more reliable risk-aware execution through selective abstention, with a corresponding trade-off between grasp reliability and harvesting yield.