LAVAPilot: Lightweight UAV Trajectory Planner with Situational Awareness for Embedded Autonomy to Track and Locate Radio-tags

Hoa Van Nguyen, Fei Chen, Joshua Chesser, Hamid Rezatofighi, Damith RanasingheIROS 2020 · DOI · OpenReview

TL;DR

This paper introduces LAVAPilot, a novel, computationally efficient trajectory planning system for autonomous drones to track and locate radio-tagged wildlife. It achieves this by prioritizing proximity to reduce measurement uncertainty while simultaneously ensuring a safe distance from animals, a crucial aspect for conservation efforts.

What Problem Does This Paper Address?

Tracking and locating wildlife using radio tags is vital for conservation but is traditionally a manual, time-consuming, and labor-intensive process. While drones offer a promising solution for automating this task, existing approaches often struggle with the limited computational power and battery life of onboard drone systems. Furthermore, a critical challenge is avoiding disturbance to the wildlife, as drones flying too close can agitate or scare animals, compromising data collection and animal welfare. This paper tackles the problem of developing a fully autonomous drone system that can efficiently track multiple radio-tagged animals, perform real-time planning onboard, and crucially, maintain a safe distance from the wildlife.

Methods

The core innovation is the LAVAPilot algorithm, a lightweight, task-based trajectory planner. Unlike traditional information-gain-based methods that can be computationally expensive, LAVAPilot leverages a simpler, greedy evaluation strategy. It's guided by two key intuitions:

  • Proximity reduces uncertainty: Moving closer to a radio tag generally leads to stronger and less noisy signal measurements, thus reducing position estimation uncertainty. LAVAPilot prioritizes moving towards objects with the lowest current uncertainty.
  • Situational awareness: To avoid disturbing wildlife, LAVAPilot incorporates a "void probability functional." This function ensures the drone maintains a predefined safe distance ($r_{min}$) from all objects of interest.

The system combines this planner with a sequential importance resampling particle filter for tracking the radio-tagged objects. The entire system is designed for embedded autonomy, meaning all planning and tracking computations are performed onboard the drone, eliminating the need for a cumbersome ground control station.

Datasets

The paper does not explicitly mention or introduce new datasets for training. Instead, it relies on:

  • Extensive Monte Carlo simulations: These simulations involve scenarios with multiple mobile radio-tagged objects in a defined search area.
  • Field experiments: The authors conducted six fully autonomous flights using a quad-copter UAV equipped with VHF radio tags. These experiments were performed on university-owned property.

Key Insights and Ideas

  • Simplicity can be powerful: The paper demonstrates that a simpler, intuition-driven planning strategy (moving closer to reduce uncertainty) can be as effective as complex information-gain methods for localization, especially when computational resources are limited.
  • "Void" for wildlife safety: The void probability functional is a clever way to embed "situational awareness" into the planning process, ensuring the drone's actions are considerate of the animals' well-being. This is a crucial step towards ethical and effective wildlife monitoring with drones.
  • Embedded autonomy is achievable: The successful implementation and validation of LAVAPilot on a Raspberry Pi 3 demonstrate that sophisticated autonomous tasks like joint tracking and planning can be performed onboard resource-constrained aerial robots.
  • Trade-offs are manageable: While LAVAPilot might take slightly longer to complete a mission compared to some information-gain methods, the significant reduction in computational cost and the guaranteed safety margin make it a highly practical solution.

Key Results

  • Computational Cost Reduction: LAVAPilot reduces the computational cost of planning by 98.5% compared to information-gain-based methods (specifically Shannon entropy).
  • Real-time Performance: The algorithm achieves real-time planning on an embedded computer (Raspberry Pi 3) with a maximum planning time of 0.1 seconds.
  • Localization Accuracy: LAVAPilot achieves similar localization accuracy (measured by RMS error) to information-gain-based methods, even with the void constraint applied.
  • Mission Completion Time: While computational cost is reduced, the mission flight time is slightly longer (around 17% increase compared to Rényi) due to the void constraint potentially limiting some direct paths.
  • Field Validation: Field experiments confirmed that LAVAPilot effectively navigates the UAV around radio tags while maintaining a safe distance, validating the void functional's effectiveness.

Limitations and Future Work

  • Assumptions: The current formulation assumes a known number of objects and does not explicitly handle missed detections or false alarms. The simulation also assumes uniform transmitted power for all radio tags, which might not hold true in real-world scenarios.
  • Future Work: The authors suggest exploring other filtering algorithms, such as Box-particle filters, to potentially improve tracking accuracy and reduce computational time further. They also highlight the potential for extending this work to more complex scenarios with dynamic environments and a larger number of targets.

Summary updated

Abstract

Tracking and locating radio-tagged wildlife is a labor-intensive and time-consuming task necessary in wildlife conservation. In this article, we focus on the problem of achieving embedded autonomy for a resource-limited aerial robot for the task capable of avoiding undesirable disturbances to wildlife. We employ a lightweight sensor system capable of simultaneous (noisy) measurements of radio signal strength information from multiple tags for estimating object locations. We formulate a new lightweight task-based trajectory planning method-LAVAPilot-with a greedy evaluation strategy and a void functional formulation to achieve situational awareness to maintain a safe distance from objects of interest. Conceptually, we embed our intuition of moving closer to reduce the uncertainty of measurements into LAVAPilot instead of employing a computationally intensive information gain based planning strategy. We employ LAVAPilot and the sensor to build a lightweight aerial robot platform with fully embedded autonomy for jointly tracking and planning to track and locate multiple VHF radio collar tags used by conservation biologists. Using extensive Monte Carlo simulation-based experiments, implementations on a single board compute module, and field experiments using an aerial robot platform with multiple VHF radio collar tags, we evaluate our joint planning and tracking algorithms. Further, we compare our method with other information-based planning methods with and without situational awareness to demonstrate the effectiveness of our robot executing LAVAPilot. Our experiments demonstrate that LAVAPilot significantly reduces (by 98.5%) the computational cost of planning to enable real-time planning decisions whilst achieving similar localization accuracy of objects compared to information gain based planning methods, albeit taking a slightly longer time to complete a mission. To support research in the field, and conservation biology, we also open source the complete project. In particular, to the best of our knowledge, this is the first demonstration of a fully autonomous aerial robot system where trajectory planning and tracking to survey and locate multiple radio-tagged objects are achieved onboard.
BibTeX
@inproceedings{nguyen2020lavapilot,
  author = {Hoa Van Nguyen and Fei Chen and Joshua Chesser and Hamid Rezatofighi and Damith Ranasinghe},
  title = {{LAVAPilot: Lightweight UAV Trajectory Planner with Situational Awareness for Embedded Autonomy to Track and Locate Radio-tags}},
  booktitle = {IEEE/RSJ International Conference on Intelligent Robots and Systems},
  year = {2020},
  doi = {10.1109/iros45743.2020.9341615},
}
Figure