Autonomous Exploration In Unknown Urban Environments For Unmanned Aerial Vehicles

David Hyunchul Shim, Hoam Chung, H. Jin Kim, S. Shankar SastryAIAA 2005 · DOI

TL;DR

This paper presents a novel approach for Unmanned Aerial Vehicles (UAVs) to autonomously explore unknown urban environments. By combining real-time local obstacle mapping with Model Predictive Control (MPC), the system can dynamically replan flight paths to avoid collisions, enabling safer and more capable urban navigation.

What Problem Does This Paper Address?

As UAVs are increasingly tasked with complex missions in urban settings, the need for autonomous navigation in cluttered and unknown environments is paramount. Traditional methods often rely on pre-programmed maps or human operators, which are insufficient for dynamic, unpredictable urban canyons. The core challenge is enabling a UAV to simultaneously build a map of its surroundings and plan a collision-free path through it, all in real-time, while operating under the constraints of limited onboard payload and high-speed flight.

Methods

The paper proposes an integrated approach centered around two key components:

  • Local Obstacle Map Building: An onboard laser scanner (specifically, a SICK LMS-200) is used to gather distance measurements of nearby obstacles. This raw data is then processed through a series of coordinate transformations (from scanner-body to vehicle-body to spatial coordinates) to create a local, real-time map of the environment. A FIFO buffer is used to store recent measurements, with filtering to discard noise and small, non-threatening objects.
  • Model Predictive Control (MPC) for Trajectory Generation: A nonlinear MPC (NMPC) framework is employed to generate a collision-free reference trajectory. The MPC algorithm uses a simplified vehicle dynamics model and a cost function that penalizes proximity to the nearest obstacle (using a potential field concept). This allows for simultaneous trajectory generation and obstacle avoidance. The MPC replans the path in real-time, generating a sequence of desired velocities that are then sent to the UAV's lower-level position tracking system.

The novelty lies in the combination of these two elements for urban exploration, specifically the real-time integration of laser scanner data into an MPC framework for dynamic path replanning.

Datasets

Not discussed in the paper. The experiments were conducted in a simulated urban environment using portable canopies and in a controlled field experiment setup.

Key Insights and Ideas

  • MPC as a Unified Framework: MPC proves effective not only for stabilization but also for integrating complex tasks like trajectory generation and obstacle avoidance into a single optimization problem.
  • Real-time Local Mapping is Crucial: Relying solely on pre-programmed maps is insufficient for unknown urban environments. The ability to build and update a local map dynamically from sensor data is essential.
  • Potential Field for Obstacle Avoidance: Incorporating a potential field term into the MPC cost function provides a robust way to guide the UAV away from obstacles, effectively acting as a "repulsive force."
  • Balancing Exploration and Safety: The MPC's cost function allows for tuning parameters (like $K_{obs}$) to balance the desire to follow a reference path with the need to break away and avoid obstacles.
  • Simplified Dynamics for Real-time Performance: Using a simplified vehicle dynamics model for the MPC trajectory generation reduces computational load, making real-time operation feasible within the UAV's constraints.

Key Results

The proposed approach was validated through simulations and experiments using a Berkeley UAV.

  • Successful Collision-Free Navigation: The system demonstrated the ability to generate collision-free trajectories and successfully guide the UAV around simulated buildings (canopies) with minimal safe clearance.
  • Real-time Performance: The MPC algorithm, implemented in C, was configured to run at a 10Hz update rate, enabling responsive navigation.
  • Laser Scanner Effectiveness: The onboard laser scanner proved capable of accurately detecting simulated urban obstacles within its line of sight.

Limitations and Future Work

  • Limitations:
    • The current implementation uses a simplified vehicle dynamics model for trajectory generation, which might not capture all nuances of the full vehicle dynamics.
  • The accuracy of the spatial obstacle coordinates depends on the accuracy of the vehicle's position and attitude estimation (INS/GPS).
  • The laser scanner has limitations in range and field of view, requiring careful mounting and potentially multiple sensors for full coverage.
  • The experiments were conducted in a simulated urban environment and a controlled field, not a fully uncontrolled, complex urban setting.
  • Future Work:
    • The paper suggests extending the MPC framework to handle more complex scenarios, potentially including dynamic obstacles and more sophisticated exploration strategies.
  • Further research could explore more advanced map-building techniques and sensor fusion to improve environmental understanding.
  • The authors also mention the potential for decentralized MPC for multi-UAV coordination, which could be a future direction.

Summary updated

Abstract

§In this paper, we present an autonomous exploration method for unmanned aerial vehicles in unknown urban environment. We address two major aspects of explorationgathering information about the surroundings and avoiding obstacles in the flight path- by building local obstacle maps and solving for confli ct-free trajectory using model predictive control (MPC) framework. For obstacle sensing, an onboard laser scanner is used to detect nearby objects around the vehicle. An MPC algorithm with a cost function that penalizes the proximity to the nearest obstacle replans the fligh t path in real-time. The adjusted trajectory is sent to the position tracking layer in the UAV a vionics. The proposed approach is implemented on Berkeley rotorcraft UAVs and successfully tested in a series of flights in urban obstacle setup.
BibTeX
@inproceedings{shim2005autonomous,
  author = {David Hyunchul Shim and Hoam Chung and H. Jin Kim and S. Shankar Sastry},
  title = {{Autonomous Exploration In Unknown Urban Environments For Unmanned Aerial Vehicles}},
  booktitle = {AIAA},
  year = {2005},
  doi = {10.2514/6.2005-6478},
}
Figure