2. Concept Overview
• Autonomous Path-Planning for Optimal Remote Structural Inspection
• 3D reconstruction of the infrastructure facilities
• Damage detection
• Non-destructive Testing - Contact-based Inspection
• Infrastructure Maintenance via Aerial Manipulation
Ultimate goal: Complete Autonomous Infrastructure Inspection and
Maintenance Solution
Illustration of Infrastructure Inspection Scenario
3. Concept Overview
A challenging mission!
• Self-localization within complex environments
• Robust and safe flight control
• Efficient path-planning for remote inspection
• Visual/Inertial-based environmental reconstruction
• Obstacle Avoidance
• Physical Interaction Control for contact-based inspection
• Aerial robotic manipulation
A great scientific/technological challenge with significant financial impact!
Illustration of Infrastructure Inspection Scenario
4. Presentation Outline
Autonomous Path-Planning for Remote Infrastructure Inspection
3D reconstruction of the environment
State estimation and mapping
Stereo-IMU Dense Reconstruction
Physical Interaction Control for Contact-based Inspection
Aerial robotic manipulation towards Infrastructure Maintenance
5. Path-planning for Infrastructure Inspection
Develop algorithms that can:
• Compute the path that an aerial vehicle has
to follow to fully inspect a 3D structure
assuming a specific visibility sensor model.
• Compute such a path with minimum distance
and account for vehicle limitations and
constraints.
• Compute inspection path in reasonable time
despite its strong ND-hardness
• Account for local changes of the structure and
have robustness against uncertainty
6. Path-planning for Infrastructure Inspection
Create a meta-tree structure to enable an incremental sampling-based approach to
inspection path-planning
Problem: given a representation
of the structure, compute the
optimal inspection path.
Challenge: ensure full coverage
and cost optimality while
handling constraints and keeping
the computational times low.
Challenge: Have a good solution
at “anytime”.
Challenge: handle uncertainty.
Goal: an efficient “anytime”
inspection planning algorithm
with performance guarantees.
7. Path-planning for Infrastructure Inspection
Evaluation in a 2D
scenario for Holonomic
and Nonholonomic
vehicles.
Convergence plots are
very promising!
8. Path-planning for Infrastructure Inspection
Experimentally Evaluated considering both
holonomic and nonholonomic vehicle trajectories.
First paths are computed in a few seconds.
10. State Estimation & Dense Reconstruction
Fundamental requirement: Deployment of robust field robots capable of self-localization and
reconstructing the model of the environment.
Photo: Mine Inspection scenario
M. Achtelik, M. Burri, M. Chli, P. Furgale, S. Leutenegger, S. Lynen, J. Nikolic, S. Omari, J. Rehder
14. Aerial Contact-based Inspection
Fully automate the process of contact-based inspection at all technologies involved:
• Control
• Perception
• Planning
This task is combined with the task of remote
inspection as the selection of the “inspection
through contact points” are based on the
previously derived model of the environment.
15. Aerial Contact-based Inspection
G. Darivianakis, K. Alexis, M. Burri, R. Siegwart, "Hybrid Predictive
Control for Aerial Robotic Physical Interaction towards Inspection
Operations", IEEE International Conference on Robotics and
Automation, ICRA 2014, Hong Kong, China (accepted)
16. Aerial Contact-based Inspection
Subject to:
• Hybrid vehicle dynamics capturing all operation modes
• Encoded Input/State constraints
• Obstacle polyhedric constraints
Hybrid Model Predictive Control problem over a prediction horizon N
18. Aerial Contact-based Inspection
Problem: Compute the optimal
route that inspects (in the
sense of visiting) all the points
in the structure.
Challenge: close to real-time
solvers.
Challenge: handle non-smooth
surfaces
Challenge: allow for “on-the-
fly” variation of the contact
points
20. Heavy-Aerial Robotic Work-Task Execution
Employ a tilt-rotor configuration to enable the
execution of work-tasks that require forces
exceeding the capabilities of typical MAVs.
C. Papachristos et. al.
22. Summary
Aerial Robots can greatly benefit Infrastructure Inspection and Maintenance
It is a task that poses Great Scientific Challenges
Main Areas of Scientific Research
It corresponds to a market with great potential
Single and Multi-robot Path-Planning for Inspection
Single and Multi-robot Perception, Localization and Mapping
Flight Control for Physical Interaction and Manipulation
Aerial Manipulation
Good time to make these systems autonomous and reliable – good time for business
Aerial robot navigation and obstacle avoidance
23. Short list of References
[1] G. Darivianakis, K. Alexis, M. Burri, R. Siegwart, "Hybrid Predictive Control
for Aerial Robotic Physical Interaction towards Inspection Operations“, IEEE
International Conference on Robotics and Automation (ICRA), 2014, Hong
Kong, China
[2] C. Papachristos, K. Alexis, A. Tzes, " Efficient Force Exertion for Aerial Robotic
Manipulation: Exploiting the Thrust-Vectoring Authority of a Tri-TiltRotor UAV
"IEEE International Conference on Robotics & Automation (ICRA), 2014,
Hong Kong, China
[3] K. Alexis, C. Huerzeler, R. Siegwart, “Hybrid Modeling and Control of a Coaxial
Unmanned Rotorcraft Interacting with its Environment through Contact”, IEEE
International Conference on Robotics and Automation, ICRA 2013, Karlsruhe,
Germany
[4] S. Leutenegger et.al., “Keyframe-Based Visual-Inertial SLAM Using Nonlinear
Optimization”, RSS 2013
[5] Markus Achtelik, Sammy Omari, Roland Siegwart, Closing the MAV navigation
loop: from control to localization and path-planning, IROS 2013
…and on going work…