Autonomous 3D Environment Exploration/Inspection

Project Overview

  • We study the 3D environment exploration and inspection problem. A current application is the autonomous pipeline inspection.
  • Complicated pipeline environment can be inspected using a robot equipped with cameras, controlled and monitored by an experienced inspector. Human works may then avoid getting into the hazadous pipeline environment for the inspection. However, remote control and online monitoring can still be labor-intensive and time consuming, a fully autonomous inspection system that can recognize the dynamic surrounding environment and perform inspection is highly desirable.
  • A fundamental open problem for autonomous inspection: how to conduct the most efficient yet thorough inspection? More specifically, given an arbitrarily complex 3D environment for the robot to inspect, how many inspection spots are necessary to visually cover the entire surrounding region completely?
  • This problem is related to a geometric problem: the 3D Guarding Problem.
  • By efficiently finding approximate solutions of the NP-hard 3D guarding problem and apply them to the autonomous pipeline inspection system, we can detect on very few points yet thoroughly covers the entire environment. This can greatly benefit the autonomous design of inspection and exploration robots.

  • Project Members

  • Dr. Xin Li
  • Wuyi Yu
  • Collaborators

  • Xiamen University: Prof. Maoqing Li
  • Florida International University: Prof. S. S. Iyengar
  • Recent Work

    The Prototype Robot FAMPER

    (a) (b): The design of FAMPER;
    (c) (d): FAMPER's is inspecting in a pipe.

    The Simulation Environment

    (a) A 3D model of the FAMPHER robot;
    (b-d) The FAMPHER robot is inspecting different pipelines; red points are the optimal inspection points we computed.


    2. Wuyi Yu, Maoqing Li and Xin Li. "Optimizing Pyramid Visibility Coverage for Autonomous Robots in 3D Environment," Control and Intelligent System, vol. 42, pp. 9-15, 2014.. [PDF] [Bibtex]

    This paper studies the optimal visibility coverage for autonomous robots in complex 3D environments. When a robot equipped with a range sensor is sent to inspect a 3D region, we want the complete visual coverage on the entire region using smallest number of scans. The practical sensor equipped on the robot usually has a pyramid-shaped visible range with restricted scanning distance and angle. Finding the optimal pyramid visibility coverage of a 3D region is NP hard. In this paper we present an efficient computation algorithm to find a good approximate solution. Our framework allows the user to flexibly specify a coverage rate parameter to balance the percentage of visibility and the required guarding points for the given region. The algorithm is assessed in a simulated 3D pipeline environment for the detection of leak, clog, and deformation.

    1. Xin Li, Wuyi Yu, Xiao Lin, and S. S. Iyengar. "On Optimizing Autonomous Pipeline Inspection in 3D Environment," IEEE Trans. on Robotics (TRO), vol. 28, no. 1, pp. 223-233, 2012.. [PDF] [Video] [Bibtex]

    This paper introduces a hierarchical optimization algorithm to an open NP-hard 3D guarding problem for massive data sets. The proposed hierarchical integer linear programming (HILP) algorithm can find the fewest spots necessary to cover an entire given 3D region. Efficiently solving this problem can greatly benefit autonomous pipeline monitoring and inspections. Unlike most existing pipeline inspection systems that focus on designing mobility and control of the explore robots, this frame- work focuses on planning automatic and thorough inspection in a complex environment. We demonstrate its efficacy on a simulated system built upon scanned pipelines environments using our prototype robots, in which leaks, clogs, and deformation can be thoroughly detected.