Aggressive Quadrotor Flight through Narrow Gaps with Onboard Sensing and Computing using Active Vision (1612.00291v6)

Abstract: We address one of the main challenges towards autonomous quadrotor flight in complex environments, which is flight through narrow gaps. While previous works relied on off-board localization systems or on accurate prior knowledge of the gap position and orientation, we rely solely on onboard sensing and computing and estimate the full state by fusing gap detection from a single onboard camera with an IMU. This problem is challenging for two reasons: (i) the quadrotor pose uncertainty with respect to the gap increases quadratically with the distance from the gap; (ii) the quadrotor has to actively control its orientation towards the gap to enable state estimation (i.e., active vision). We solve this problem by generating a trajectory that considers geometric, dynamic, and perception constraints: during the approach maneuver, the quadrotor always faces the gap to allow state estimation, while respecting the vehicle dynamics; during the traverse through the gap, the distance of the quadrotor to the edges of the gap is maximized. Furthermore, we replan the trajectory during its execution to cope with the varying uncertainty of the state estimate. We successfully evaluate and demonstrate the proposed approach in many real experiments. To the best of our knowledge, this is the first work that addresses and achieves autonomous, aggressive flight through narrow gaps using only onboard sensing and computing and without prior knowledge of the pose of the gap.

• The paper presents a method for autonomous quadrotor flight through narrow gaps using only onboard camera, IMU, and computing for state estimation and trajectory generation.
• The approach utilizes a two-phase trajectory generation algorithm that considers geometric, dynamic, and perception constraints, achieving an 80% success rate in real-world experiments.
• This work has implications for autonomous navigation in obstacle-laden environments like search and rescue, with potential future enhancements in sensors or integration of machine learning.

Aggressive Quadrotor Flight through Narrow Gaps with Onboard Sensing and Computing using Active Vision

The paper "Aggressive Quadrotor Flight through Narrow Gaps using Onboard Sensing and Computing with Active Vision" presents an advanced methodology for autonomous quadrotor navigation in complex environments. The authors tackle the intricate challenge of maneuvering quadrotors through narrow gaps without relying on external localization systems or prior knowledge of gap positioning. The solution exploits onboard sensing and computing, utilizing a single forward-facing camera in conjunction with an IMU to estimate the quadrotor's full state.

Overview of the Approach

This research identifies two critical obstacles in achieving autonomous flight through narrow gaps: the increasing pose uncertainty of the quadrotor relative to the gap with distance, and the necessity for active orientation control to facilitate state estimation (active vision). To address these, the authors propose a trajectory generation algorithm that integrates geometric, dynamic, and perception constraints.

The algorithm is divided into two key phases:

1. Traverse Trajectory: This trajectory ensures the quadrotor path intersects the center of the gap, minimizing collision risks by maintaining a trajectory within a plane orthogonally aligned to the gap. Calculations consider both dynamic feasibility and maximum allowable velocities, optimizing the trajectory pre-execution.
2. Approach Trajectory: The quadrotor approaches the gap while constantly facing it, enabled by selecting appropriate yaw angles to enhance visual state estimation. Trajectories are rapidly computed onboard, allowing for continuous replanning to accommodate varying uncertainties until the quadrotor executes the traverse.

Experimental Validation

The system was rigorously tested through numerous real-world experiments with variable gap orientations. The quadrotor reached angular velocities up to 400 degrees per second and achieved a high success rate of 80%. Failures were primarily attributed to extended periods of missing detections necessary for pose estimation, particularly at steep angles of the gap or high velocities causing motion blur.

Implications and Future Directions

This work advances autonomous system capabilities in obstacle-laden environments, particularly useful in scenarios such as search and rescue operations where external systems may be unavailable or infeasible. The success of this approach underpins its potential applicability to a broad range of autonomous vehicle tasks where onboard computation and sensing must prevail over external configurations.

Future research might explore enhancements in sensor technology to extend capabilities with lower error margins, potentially incorporating newer camera technologies or alternative sensing modalities to improve reliability under high dynamic maneuvers or varied environmental conditions. Additionally, machine learning algorithms could be integrated to better handle partial or noisy data, refining state estimation under adverse conditions.

In summary, this paper demonstrates substantial progress in autonomous quadrotor navigation capabilities, showcasing a method that leverages onboard computing power to navigate complex environments with significant agility and precision. As AI and sensing technologies evolve, such methodologies are expected to become increasingly sophisticated, offering robust solutions to dynamic real-world challenges.