Autonomous Mission Management for Forest Search with Multiple Unmanned Aerial Vehicles

Kemao Peng, Feng Lin, Ben M. ChenICINCO 2016 · DOI

TL;DR

This paper introduces an enhanced Hierarchical-Distributed Methodology (HDM) for coordinating multiple Unmanned Aerial Vehicles (UAVs) to autonomously search a forest. The system effectively decomposes complex missions into manageable tasks, allowing for efficient collaboration and adaptation even when individual UAVs encounter issues.

What Problem Does This Paper Address?

Coordinating multiple UAVs for complex missions, like searching a large, potentially hazardous area, presents significant challenges. Traditional approaches often struggle with:

  • Event-Driven Complexity: Hybrid systems, which combine continuous dynamics (like flight) with discrete events (like task completion), are difficult to model and control. The sheer number of potential events can make designing robust control laws impractical.
  • Task Decomposition and Assignment: Breaking down a large mission into smaller, assignable tasks for individual UAVs, and then managing their progress, is a complex logistical problem.
  • Dynamic Re-scheduling: When unforeseen events occur (e.g., a UAV failure), the system needs to adapt its plan dynamically without compromising the overall mission objective.
  • UAV Capabilities: UAVs can only execute dynamic commands (like setting a velocity or trajectory). Event-driven commands need to be translated into these executable dynamic commands, considering the UAV's physical limitations.

This paper aims to address these issues by proposing a structured and adaptable framework for autonomous mission management in multi-UAV systems.

Methods

The core of the paper is the Enhanced Hierarchical-Distributed Methodology (HDM). This approach builds upon existing HDM concepts by introducing:

  • Hierarchical Control: The system operates at multiple levels. A high-level mission management system (AMM) oversees the overall mission, while lower-level controllers handle individual UAV actions.
  • Event Control Law Partitioning: The event control law is split into two parts:
  • Group Event Control Law: This law coordinates the entire group of UAVs, focusing on mission decomposition, scheduling, and overall progress. A designated "leader" UAV manages this.
  • Individual Event Control Laws: These laws manage the tasks and subtasks assigned to each individual UAV, ensuring they can execute their specific roles.
  • Mission Decomposition: Missions are broken down into a series and/or parallel sequence of "submissions" (high-level goals) and then further into "tasks" and "subtasks." This allows for flexible task allocation and execution.
  • Dynamic Re-scheduling: The system is designed to handle situations where UAVs might be lost. The remaining UAVs can adapt and re-schedule tasks, as demonstrated in the simulation.
  • Event Command to Dynamic Command Conversion: A crucial aspect is how abstract event commands (e.g., "search this area") are translated into concrete, executable dynamic commands (e.g., specific flight paths, velocities, and accelerations) that the UAVs can follow. This conversion considers phases like acceleration, velocity holding, deceleration, and hovering.

Datasets

Not discussed in the paper. The paper focuses on a simulated forest search scenario.

Key Insights and Ideas

  • Structured Decomposition is Key: The hierarchical decomposition of missions into submissions, tasks, and subtasks is fundamental to managing complexity and enabling parallel execution. This allows for a more perfect decomposition of missions.
  • Leader-Follower Dynamics: The concept of a designated leader (which can change) for each batch of UAVs simplifies group coordination and data aggregation.
  • Adaptability through Online Scheduling: The ability to schedule the third batch of UAVs online, based on the results and potential failures of the first two batches, is a critical feature for robust mission execution in dynamic environments.
  • Bridging the Event-Dynamic Gap: The explicit conversion of event commands to dynamic commands, considering flight phases, is essential for making abstract mission goals executable by physical UAVs.
  • Balancing Event Granularity: The paper emphasizes the importance of defining a reasonable number of event variables to avoid making control laws unmanageable, a common challenge in hybrid systems.

Key Results

The paper demonstrates the effectiveness of the proposed Enhanced HDM through a simulation of nine UAVs searching a 500x500 meter forest field.

  • Successful Mission Completion: The simulation results show that the designed AMM system successfully coordinated the group of UAVs to search the forest field collaboratively.
  • Handling Failures: The simulation included scenarios where UAVs were lost in the first and second batches. The system adapted by re-scheduling and assigning tasks to remaining UAVs, ultimately ensuring the entire forest was searched.
  • Efficient Data Aggregation: The leader UAVs effectively collected, merged, and processed 2D maps and detected targets from their respective batches.
  • Verifiable Implementation: The authors state that the AMM system was successfully verified in a high-fidelity simulator.

Limitations and Future Work

  • Accident Handling: While the paper demonstrates adaptation to UAV loss, the authors note that "the mission reschedule is to be studied when an accident happens." This suggests that more complex accident scenarios and their rescheduling implications warrant further investigation.
  • Communication Constraints: The simulation assumes limited communication between UAVs and the ground station, with communication primarily occurring at the beginning and end of processes. Real-world scenarios might involve more complex communication needs and constraints.
  • Real-World Deployment: The current work is based on simulation. Future work would involve testing and validating the system on actual UAV hardware in realistic environments.

Summary updated

BibTeX
@inproceedings{peng2016autonomous,
  author = {Kemao Peng and Feng Lin and Ben M. Chen},
  title = {{Autonomous Mission Management for Forest Search with Multiple Unmanned Aerial Vehicles}},
  booktitle = {ICINCO},
  year = {2016},
  doi = {10.5220/0005949603850392},
}
Figure