Nuclear Environments Inspection with Micro Aerial Vehicles: Algorithms and Experiments

Dinesh Thakur, Giuseppe Loianno, Wenxin Liu, Vijay Kumar

Abstract In this work, we address the estimation, planning, control and mapping problems to allow a small quadrotor to autonomously inspect the interior of hazardous damaged nuclear sites. These algorithms run onboard on a computationally limited CPU. We investigate the effect of varying illumination on the system performance. To the best of our knowledge, this is the first fully autonomous system of this size and scale applied to inspect the interior of a full scale mock-up of a Primary Containment Vessel (PCV). The proposed solution opens up new ways to inspect nuclear reactors and to support nuclear decommissioning, which is well known to be a dangerous, long and tedious process. Experimental results with varying illumination conditions show the ability to navigate a full scale mock-up PCV pedestal and create a map of the environment, while concurrently avoiding obstacles.

Introduction: The Challenge of Nuclear Decommissioning

Decommissioning damaged nuclear reactors, such as those at the Fukushima Daiichi site, presents a monumental challenge for robotics. These environments are highly radioactive, cluttered, and confined, often requiring navigation through access hatches as small as 0.1 to 0.3 meters in diameter. Current robotic solutions typically involve large ground or aerial vehicles, ranging from 1 to 10 meters in size and weighing up to 1000 kg. These systems rely heavily on remote teleoperation and continuous communication links, which are prone to failure in shielded, complex structures. Furthermore, the lack of GPS and poor lighting conditions make traditional autonomous navigation impossible.

Diagram of a Primary Containment Vessel (PCV) and the corresponding full-scale experimental mock-up used for testing.

To address these limitations, this research introduces a palm-sized (0.16 m diameter), 236 g quadrotor designed for full autonomy. This Micro Aerial Vehicle (MAV) is small enough to penetrate narrow entry points and capable of performing state estimation, mapping, and obstacle avoidance entirely onboard, without the need for external computation or human intervention.

System Architecture and Methodology

The MAV is built around the Qualcomm Snapdragon Flight board, which integrates an ARM processor and a Digital Signal Processor (DSP). The software stack is optimized to run concurrently on this computationally constrained hardware:

  • State Estimation & Control: The system utilizes Visual Inertial Odometry (VIO), fusing data from an onboard IMU with a downward-facing VGA camera featuring a 170° field of view. An Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) process this data at 500 Hz to provide high-frequency state estimates for a nonlinear proportional-derivative controller.
  • Mapping: A front-facing stereo camera generates a dense 3D voxel grid occupancy map. To manage CPU load, image rectification is multi-threaded, and a block-matching algorithm extracts disparity maps at 15 Hz to identify obstacles in the flight path.
  • Reactive Planning: While the drone follows a pre-defined global trajectory, it employs an online reactive planner to handle unexpected obstacles. Using Quintic B-splines, the system dynamically recalculates smooth trajectories that ensure continuous velocity and acceleration while deviating around hazards.

Experimental Validation

The system was rigorously tested in a full-scale, 5-meter diameter mock-up of a Primary Containment Vessel (PCV) pedestal at the Southwest Research Institute (SwRI). These tests aimed to replicate the cluttered and dimly lit conditions found in actual damaged reactors.

Interior view of a damaged nuclear environment, illustrating the extreme clutter and challenging lighting conditions that the MAV must navigate.

Obstacle Avoidance and Inspection

During end-to-end inspection trials, the MAV successfully navigated from outside the mock-up into the cluttered pedestal area. It detected and avoided unmapped obstacles, including PVC pipes and aluminum trusses as small as 0.25 m in diameter, while maintaining a nominal flight velocity of 0.5 m/s. The stereo vision system proved accurate enough to map the environment in real-time despite the micro-drone's small camera baseline.

Luminance Stress Tests

Given the reliance on visual features for VIO, the authors conducted stress tests under varying light levels. The system maintained stable flight at 41.8 lux and 9.5 lux. However, at 0.5 lux, the degradation of visual features led to significant VIO drift and a subsequent crash. This result highlights the critical need for integrated active lighting in future deployments.

Insights and Future Work

This research demonstrates that a complete autonomy stack can be successfully deployed on a sub-250g platform using mobile-grade CPUs. Key insights include the viability of stereo vision for micro-scale obstacle avoidance and the identification of ambient lighting as a primary limiting factor for VIO-based systems.

Future work will focus on integrating an onboard LED payload for active lighting, implementing a real-time global planner to avoid local minima in complex mazes, and further miniaturizing the platform to a 0.1 m diameter. These advancements will move the technology closer to deployment in real-world nuclear decommissioning missions, reducing the risk to human personnel.

Summary updated

Abstract

In this work, we address the estimation, planning, control and mapping problems to allow a small quadrotor to autonomously inspect the interior of hazardous damaged nuclear sites. These algorithms run onboard on a computationally limited CPU. We investigate the effect of varying illumination on the system performance. To the best of our knowledge, this is the first fully autonomous system of this size and scale applied to inspect the interior of a full scale mock-up of a Primary Containment Vessel (PCV). The proposed solution opens up new ways to inspect nuclear reactors and to support nuclear decommissioning, which is well known to be a dangerous, long and tedious process. Experimental results with varying illumination conditions show the ability to navigate a full scale mock-up PCV pedestal and create a map of the environment, while concurrently avoiding obstacles.
BibTeX
@inproceedings{thakur2018nuclear,
  author = {Dinesh Thakur and Giuseppe Loianno and Wenxin Liu and Vijay Kumar},
  title = {{Nuclear Environments Inspection with Micro Aerial Vehicles: Algorithms and Experiments}},
  booktitle = {ISER},
  year = {2018},
  url = {https://openreview.net/forum?id=LZ4NYoMTgU},
}
Figure